[{"data":1,"prerenderedAt":2043},["ShallowReactive",2],{"blog-posts":3},[4,388,633,782,993,1183,1301,1472,1646,1888],{"id":5,"title":6,"author":7,"body":8,"date":375,"description":376,"extension":377,"meta":378,"navigation":379,"path":380,"seo":381,"stem":382,"tags":383,"__hash__":387},"blog\u002Fblog\u002Fan-n-of-1-study.md","Your life is an n-of-1 study","The Daynostic Team",{"type":9,"value":10,"toc":366},"minimark",[11,18,21,26,42,45,49,61,64,68,71,83,100,112,116,128,131,135,138,191,199,205,211],[12,13,14],"blockquote",{},[15,16,17],"p",{},"The average person doesn't exist.",[15,19,20],{},"Population studies measure thousands of people to find what is true on average. The average is a summary of a crowd, and no single person matches it exactly. \"Does sunlight improve mood?\" is a question about the population. \"Does sunlight improve this person's mood?\" is a different question, and a harder one. Below is what the research says about weather, weekdays and mood, and why it points toward analysing a single life as an n-of-1 study: a study with one participant, run with the same care as a study with thousands.",[22,23,25],"h2",{"id":24},"the-weather-effect-is-real-small-and-individual","The weather effect is real, small and individual",[15,27,28,29,41],{},"The largest careful study of weather and mood is by Denissen and colleagues, who linked the daily diaries of 1,233 people to local weather-station records and ran a multilevel analysis.",[30,31,32],"sup",{},[33,34,40],"a",{"href":35,"ariaDescribedBy":36,"dataFootnoteRef":38,"id":39},"#user-content-fn-1",[37],"footnote-label","","user-content-fnref-1","1"," Temperature, wind and sunlight all had an effect on negative affect, but the average effect on mood was small in terms of explained variance.",[15,43,44],{},"Underneath that average, the variation between individuals was substantial, especially in the response to daylight. Some people are weather-sensitive, some are not, and a few respond in the opposite direction. The group mean blends all of them into one small number. A population study cannot tell which case applies to a given person. That person's own data can.",[22,46,48],{"id":47},"blue-monday-is-mostly-a-memory-effect","\"Blue Monday\" is mostly a memory effect",[15,50,51,52,60],{},"Mondays are widely believed to be the worst day of the week. Stone, Schneider and Harter examined day-of-week mood in roughly 340,000 people.",[30,53,54],{},[33,55,59],{"href":56,"ariaDescribedBy":57,"dataFootnoteRef":38,"id":58},"#user-content-fn-2",[37],"user-content-fnref-2","2"," Weekends and Fridays did lift mood, but the Blue Monday effect was minimal: Monday was practically indistinguishable from Tuesday, Wednesday or Thursday.",[15,62,63],{},"The gap between memory and record is the interesting result. When participants were asked to recall their worst day, 65% named Monday, a day that was no worse than midweek in their own logged data. Memory does not store the week as it happened. It stores a story about the week, and the story is shaped by expectation. A check-in logged the same evening, before the story forms, is a much better measurement.",[22,65,67],{"id":66},"why-searching-personal-data-for-patterns-goes-wrong","Why searching personal data for patterns goes wrong",[15,69,70],{},"If personal data beats both memory and population averages, mood apps should already be delivering reliable personal insights. Most do not, because searching one person's numbers for patterns runs into a well known problem in empirical science.",[15,72,73,74,82],{},"John Ioannidis described it in a widely cited 2005 paper, \"Why Most Published Research Findings Are False.\"",[30,75,76],{},[33,77,81],{"href":78,"ariaDescribedBy":79,"dataFootnoteRef":38,"id":80},"#user-content-fn-3",[37],"user-content-fnref-3","3"," A result is more likely to be false when effect sizes are small, when samples are small, and when many relationships are tested with a lot of analytical flexibility. A personal tracker meets all three conditions: small effects, a sample of one person over a few months, and dozens of variable pairs as soon as ten things are tracked.",[15,84,85,86,90,91,99],{},"When enough pairs are tested, some of them reach significance by chance. That is a property of the arithmetic, and no amount of clever software removes it. The ",[87,88,89],"em",{},"Spurious Correlations"," project illustrates it with series such as cheese consumption and deaths by bedsheet entanglement.",[30,92,93],{},[33,94,98],{"href":95,"ariaDescribedBy":96,"dataFootnoteRef":38,"id":97},"#user-content-fn-4",[37],"user-content-fnref-4","4"," An app that reports its strongest weekly correlation is mostly reporting noise.",[15,101,102,103,111],{},"The standard remedy is the Benjamini-Hochberg false-discovery-rate correction, used in genomics and clinical trials.",[30,104,105],{},[33,106,110],{"href":107,"ariaDescribedBy":108,"dataFootnoteRef":38,"id":109},"#user-content-fn-5",[37],"user-content-fnref-5","5"," It raises the threshold each result must clear in proportion to how many tests were run, which separates a finding that passed a fair test from the luckiest draw in a large set. One consequence of applying it is that an app reports nothing much more often. That is the correction working as intended.",[22,113,115],{"id":114},"n-of-1-trials-are-established-science","N-of-1 trials are established science",[15,117,118,119,127],{},"Treating a single person as the whole sample is not unscientific. N-of-1 trials, multiple-crossover experiments run within one patient, are an established and growing method in medicine.",[30,120,121],{},[33,122,126],{"href":123,"ariaDescribedBy":124,"dataFootnoteRef":38,"id":125},"#user-content-fn-6",[37],"user-content-fnref-6","6"," When a condition is stable and responds reversibly to treatment, a well-run n-of-1 trial can match or exceed a conventional group trial for that individual, because it answers the question the individual has: what works for them.",[15,129,130],{},"A daily life tracker is, in spirit, a long observational n-of-1 study. It cannot randomise the weather or blind a person to their own sleep, so it establishes association, not causation. The unit of analysis is the same, though: one life.",[22,132,134],{"id":133},"what-this-means-for-daynostic","What this means for Daynostic",[15,136,137],{},"These findings translate into a few design choices:",[139,140,141,155,167,185],"ul",{},[142,143,144,148,149],"li",{},[145,146,147],"strong",{},"Logging every evening",", because memory reconstructs. The Blue Monday result is what happens when recall is trusted over records.",[30,150,151],{},[33,152,59],{"href":56,"ariaDescribedBy":153,"dataFootnoteRef":38,"id":154},[37],"user-content-fnref-2-2",[142,156,157,160,161],{},[145,158,159],{},"Sensitivity to small, individual effects."," The weather literature shows the signal is small on average and specific to the person, so the analysis has to be able to find a small real effect without inventing one.",[30,162,163],{},[33,164,40],{"href":35,"ariaDescribedBy":165,"dataFootnoteRef":38,"id":166},[37],"user-content-fnref-1-2",[142,168,169,172,173,179],{},[145,170,171],{},"Correction for the number of tests."," Without false-discovery control, a tracker with many variables produces coincidences at a steady rate.",[30,174,175],{},[33,176,81],{"href":78,"ariaDescribedBy":177,"dataFootnoteRef":38,"id":178},[37],"user-content-fnref-3-2",[30,180,181],{},[33,182,110],{"href":107,"ariaDescribedBy":183,"dataFootnoteRef":38,"id":184},[37],"user-content-fnref-5-2",[142,186,187,190],{},[145,188,189],{},"Reporting only results that pass."," The answer to \"what is correlated this month?\" is often \"nothing yet\". An analysis that cannot give that answer is not measuring anything.",[15,192,193,194,198],{},"The ",[33,195,197],{"href":196},"\u002Fblog\u002Fhow-daynostic-decides-an-insight-is-real","companion article on the analysis pipeline"," describes what an insight goes through before it appears: rank correlations, permutation tests against shuffled history, FDR correction and day-of-week controls. The research above is the reason that pipeline exists.",[200,201,202],"tip",{},[15,203,204],{},"Population research can establish that an effect exists. Only a person's own data, tested with the same methods, can show how large it is for them.",[206,207,208],"note",{},[15,209,210],{},"Daynostic is a wellness tool, not a medical device. Its insights are statistical associations within a person's own data, not medical advice.",[212,213,216,221],"section",{"className":214,"dataFootnotes":38},[215],"footnotes",[22,217,220],{"className":218,"id":37},[219],"sr-only","Footnotes",[222,223,224,254,279,304,321,346],"ol",{},[142,225,227,228,231,232,238,239,246,247],{"id":226},"user-content-fn-1","Denissen, J. J. A., Butalid, L., Penke, L., & van Aken, M. A. G. (2008). The effects of weather on daily mood: A multilevel approach. ",[87,229,230],{},"Emotion",", 8(5), 662-667. ",[33,233,237],{"href":234,"rel":235},"https:\u002F\u002Fpubmed.ncbi.nlm.nih.gov\u002F18837616\u002F",[236],"nofollow","PubMed 18837616",". ",[33,240,245],{"href":241,"ariaLabel":242,"className":243,"dataFootnoteBackref":38},"#user-content-fnref-1","Back to reference 1",[244],"data-footnote-backref","↩"," ",[33,248,245,252],{"href":249,"ariaLabel":250,"className":251,"dataFootnoteBackref":38},"#user-content-fnref-1-2","Back to reference 1-2",[244],[30,253,59],{},[142,255,257,258,261,262,238,267,246,272],{"id":256},"user-content-fn-2","Stone, A. A., Schneider, S., & Harter, J. K. (2012). Day-of-week mood patterns in the United States: On the existence of 'Blue Monday', 'Thank God it's Friday' and weekend effects. ",[87,259,260],{},"The Journal of Positive Psychology",", 7(4), 306-314. ",[33,263,266],{"href":264,"rel":265},"https:\u002F\u002Fwww.tandfonline.com\u002Fdoi\u002Fabs\u002F10.1080\u002F17439760.2012.691980",[236],"DOI: 10.1080\u002F17439760.2012.691980",[33,268,245],{"href":269,"ariaLabel":270,"className":271,"dataFootnoteBackref":38},"#user-content-fnref-2","Back to reference 2",[244],[33,273,245,277],{"href":274,"ariaLabel":275,"className":276,"dataFootnoteBackref":38},"#user-content-fnref-2-2","Back to reference 2-2",[244],[30,278,59],{},[142,280,282,283,286,287,238,292,246,297],{"id":281},"user-content-fn-3","Ioannidis, J. P. A. (2005). Why most published research findings are false. ",[87,284,285],{},"PLoS Medicine",", 2(8), e124. ",[33,288,291],{"href":289,"rel":290},"https:\u002F\u002Fjournals.plos.org\u002Fplosmedicine\u002Farticle?id=10.1371\u002Fjournal.pmed.0020124",[236],"Full text",[33,293,245],{"href":294,"ariaLabel":295,"className":296,"dataFootnoteBackref":38},"#user-content-fnref-3","Back to reference 3",[244],[33,298,245,302],{"href":299,"ariaLabel":300,"className":301,"dataFootnoteBackref":38},"#user-content-fnref-3-2","Back to reference 3-2",[244],[30,303,59],{},[142,305,307,308,246,311,238,316],{"id":306},"user-content-fn-4","Vigen, T. ",[87,309,310],{},"Spurious Correlations.",[33,312,315],{"href":313,"rel":314},"https:\u002F\u002Ftylervigen.com\u002Fspurious-correlations",[236],"tylervigen.com\u002Fspurious-correlations",[33,317,245],{"href":318,"ariaLabel":319,"className":320,"dataFootnoteBackref":38},"#user-content-fnref-4","Back to reference 4",[244],[142,322,324,325,328,329,238,334,246,339],{"id":323},"user-content-fn-5","Benjamini, Y., & Hochberg, Y. (1995). Controlling the false discovery rate: A practical and powerful approach to multiple testing. ",[87,326,327],{},"Journal of the Royal Statistical Society: Series B",", 57(1), 289-300. ",[33,330,333],{"href":331,"rel":332},"https:\u002F\u002Fwww.jstor.org\u002Fstable\u002F2346101",[236],"JSTOR 2346101",[33,335,245],{"href":336,"ariaLabel":337,"className":338,"dataFootnoteBackref":38},"#user-content-fnref-5","Back to reference 5",[244],[33,340,245,344],{"href":341,"ariaLabel":342,"className":343,"dataFootnoteBackref":38},"#user-content-fnref-5-2","Back to reference 5-2",[244],[30,345,59],{},[142,347,349,350,355,356,238,361],{"id":348},"user-content-fn-6","\"The use of N-of-1 trials to generate real-world evidence for optimal treatment of individuals and populations\" (review). ",[33,351,354],{"href":352,"rel":353},"https:\u002F\u002Fwww.ncbi.nlm.nih.gov\u002Fpmc\u002Farticles\u002FPMC10565195\u002F",[236],"NIH\u002FPMC",". See also \"N-of-1 trials: the epitome of personalized medicine?\" in the ",[33,357,360],{"href":358,"rel":359},"https:\u002F\u002Fwww.cambridge.org\u002Fcore\u002Fjournals\u002Fjournal-of-clinical-and-translational-science\u002Farticle\u002Fnof1-trials-the-epitome-of-personalized-medicine\u002F02BB4759DBDA25620227BB149518AD58",[236],"Journal of Clinical and Translational Science",[33,362,245],{"href":363,"ariaLabel":364,"className":365,"dataFootnoteBackref":38},"#user-content-fnref-6","Back to reference 6",[244],{"title":38,"searchDepth":367,"depth":367,"links":368},2,[369,370,371,372,373,374],{"id":24,"depth":367,"text":25},{"id":47,"depth":367,"text":48},{"id":66,"depth":367,"text":67},{"id":114,"depth":367,"text":115},{"id":133,"depth":367,"text":134},{"id":37,"depth":367,"text":220},"2026-06-13","What the research says about weather, weekdays and mood: the effects are real but small on average and differ a lot between people, which is why personal analytics needs the same statistics that research uses.","md",{},true,"\u002Fblog\u002Fan-n-of-1-study",{"title":6,"description":376},"blog\u002Fan-n-of-1-study",[384,385,386],"statistics","science","research","GJLLTlHgizsqnKAR_24sHn6qEgbvL_Tv_CLbtWwDv8g",{"id":389,"title":390,"author":7,"body":391,"date":375,"description":624,"extension":377,"meta":625,"navigation":379,"path":626,"seo":627,"stem":628,"tags":629,"__hash__":632},"blog\u002Fblog\u002Fdo-habit-trackers-actually-work.md","Do habit trackers actually work?",{"type":9,"value":392,"toc":616},[393,396,400,409,412,416,419,428,431,435,438,447,450,454,463,466,470,473,512,519,522,526],[15,394,395],{},"The research gives an answer in two parts, and they point in opposite directions. The mechanism behind tracking, the act of monitoring a behavior, is among the most reliably effective results in behavioral science. The apps built on that mechanism mostly fail anyway. Both are true, and the gap between them explains most of what goes wrong with tracking apps.",[22,397,399],{"id":398},"the-mechanism","The mechanism",[15,401,402,403,408],{},"Monitoring one's own progress toward a goal is strongly supported in the research literature. The clearest evidence is a 2016 meta-analysis by Harkin and colleagues, which pooled 138 randomized studies covering nearly 20,000 people to test whether prompting people to monitor their progress changed outcomes.",[30,404,405],{},[33,406,40],{"href":35,"ariaDescribedBy":407,"dataFootnoteRef":38,"id":39},[37]," It did. Monitoring goal progress had a measurable effect on goal attainment (d+ = 0.40) across areas as varied as weight loss, smoking cessation, physical activity and medication adherence.",[15,410,411],{},"Two details from that analysis carry directly into app design. The effect was larger when progress was physically recorded, and larger when it was reported or made public. A concrete, visible record does more than a passing mental note. As far as the mechanism goes, tracking works.",[22,413,415],{"id":414},"real-world-retention","Real-world retention",[15,417,418],{},"The difficulty appears when the mechanism meets actual use. Studies measure people who, by being enrolled in a study, keep doing the thing being measured. Outside that setting, most people stop.",[15,420,421,422,427],{},"Baumel and colleagues quantified this in 2019 using independent panel data on how mental-health apps are used in practice, based on observed behavior and not self-report.",[30,423,424],{},[33,425,59],{"href":56,"ariaDescribedBy":426,"dataFootnoteRef":38,"id":58},[37]," Median 15-day retention was 3.9%, and 30-day retention was 3.3%. A month after installation, fewer than one user in twenty-five was still opening the app, and even the strongest categories (peer support, mindfulness, trackers) remained in single digits at day 30.",[15,429,430],{},"This is the main reason habit trackers fail in practice. The mechanism only operates while tracking continues, and most apps lose the user before it has time to take effect. The limiting factor is adherence. An app with a capable analysis engine that goes unopened after the first week produces the same outcome as an app that does nothing. The most important design question for a tracker is whether a person is still logging on day 60. Everything else depends on it.",[22,432,434],{"id":433},"how-long-habits-take","How long habits take",[15,436,437],{},"Surviving to day 60 is easier with a realistic picture of what happens over those weeks. The common claim of \"21 days to form a habit\" has no sound basis. The measured picture is different.",[15,439,440,441,446],{},"Lally and colleagues followed 96 people adopting a new daily behavior and recorded how long it took to become automatic.",[30,442,443],{},[33,444,81],{"href":78,"ariaDescribedBy":445,"dataFootnoteRef":38,"id":80},[37]," The median was 66 days, and the range ran from 18 to 254 days depending on the person and the behavior. Automaticity builds gradually, and one person's curve can be much slower than another's without anything being wrong.",[15,448,449],{},"One finding bears directly on how an app should handle lapses: missing a single day did not meaningfully disrupt habit formation. Streak-based designs that reset progress to zero after one missed day work against this, since a broken streak is also the point at which many people abandon the effort altogether. Punishing a missed day works against adherence, which is the variable that decides whether tracking works at all.",[22,451,453],{"id":452},"clinical-effects-of-the-apps","Clinical effects of the apps",[15,455,456,457,462],{},"A narrower question is whether a controlled trial of an actual app shows improved outcomes. The evidence here is real but modest. Linardon and colleagues meta-analyzed 66 randomized controlled trials of smartphone interventions for mental health.",[30,458,459],{},[33,460,98],{"href":95,"ariaDescribedBy":461,"dataFootnoteRef":38,"id":97},[37]," App-supported interventions produced statistically significant improvements, though generally small ones, with stronger effects when the app was combined with some human support.",[15,464,465],{},"This sets the expectation. An app promising dramatic change is claiming more than the evidence supports. What the evidence supports is a small benefit, and a small benefit sustained over months is still worth having.",[22,467,469],{"id":468},"what-the-evidence-implies","What the evidence implies",[15,471,472],{},"Taken together, the four findings point to a consistent set of design choices:",[139,474,475,484,493,502],{},[142,476,477,478,483],{},"Monitoring is most effective when the record is concrete and visible,",[30,479,480],{},[33,481,40],{"href":35,"ariaDescribedBy":482,"dataFootnoteRef":38,"id":166},[37]," so the daily entry should be something a person can see, not a value that disappears into a database. A year-at-a-glance grid is one form of a recorded, visible history.",[142,485,486,487,492],{},"Retention is low by default,",[30,488,489],{},[33,490,59],{"href":56,"ariaDescribedBy":491,"dataFootnoteRef":38,"id":154},[37]," so the cost of a daily entry has to be minimal. A check-in that takes more than about half a minute loses most users well before the 30-day mark.",[142,494,495,496,501],{},"Habits form slowly and unevenly, and occasional lapses are normal,",[30,497,498],{},[33,499,81],{"href":78,"ariaDescribedBy":500,"dataFootnoteRef":38,"id":178},[37]," so an app should absorb missed days instead of treating them as failure.",[142,503,504,505,511],{},"Effects are real but small,",[30,506,507],{},[33,508,98],{"href":95,"ariaDescribedBy":509,"dataFootnoteRef":38,"id":510},[37],"user-content-fnref-4-2"," so a tracker should not promise transformation, and should not manufacture impressive-looking results to seem effective.",[15,513,514,515,518],{},"These are the principles Daynostic is built on: a 30-second nightly check-in, a visual record of the year, no streak penalties, and an analysis layer held to the ",[33,516,517],{"href":196},"statistical standards described separately",". The mechanism belongs to the person using the app. What the app can do is lower the cost of logging and improve the odds that the person is still logging months later.",[15,520,521],{},"The mechanism is well established, and the apps usually fall short because people stop using them. A tracker's main job is to make stopping less likely.",[206,523,524],{},[15,525,210],{},[212,527,529,532],{"className":528,"dataFootnotes":38},[215],[22,530,220],{"className":531,"id":37},[219],[222,533,534,554,574,594],{},[142,535,536,537,540,541,238,546,246,549],{"id":226},"Harkin, B., Webb, T. L., Chang, B. P. I., Prestwich, A., Conner, M., Kellar, I., Benn, Y., & Sheeran, P. (2016). Does monitoring goal progress promote goal attainment? A meta-analysis of the experimental evidence. ",[87,538,539],{},"Psychological Bulletin",", 142(2), 198-229. ",[33,542,545],{"href":543,"rel":544},"https:\u002F\u002Fpubmed.ncbi.nlm.nih.gov\u002F26479070\u002F",[236],"PubMed 26479070",[33,547,245],{"href":241,"ariaLabel":242,"className":548,"dataFootnoteBackref":38},[244],[33,550,245,552],{"href":249,"ariaLabel":250,"className":551,"dataFootnoteBackref":38},[244],[30,553,59],{},[142,555,556,557,560,561,238,566,246,569],{"id":256},"Baumel, A., Muench, F., Edan, S., & Kane, J. M. (2019). Objective user engagement with mental health apps: Systematic search and panel-based usage analysis. ",[87,558,559],{},"Journal of Medical Internet Research",", 21(9), e14567. ",[33,562,565],{"href":563,"rel":564},"https:\u002F\u002Fwww.jmir.org\u002F2019\u002F9\u002Fe14567\u002F",[236],"jmir.org\u002F2019\u002F9\u002Fe14567",[33,567,245],{"href":269,"ariaLabel":270,"className":568,"dataFootnoteBackref":38},[244],[33,570,245,572],{"href":274,"ariaLabel":275,"className":571,"dataFootnoteBackref":38},[244],[30,573,59],{},[142,575,576,577,580,581,238,586,246,589],{"id":281},"Lally, P., van Jaarsveld, C. H. M., Potts, H. W. W., & Wardle, J. (2010). How are habits formed: Modelling habit formation in the real world. ",[87,578,579],{},"European Journal of Social Psychology",", 40(6), 998-1009. ",[33,582,585],{"href":583,"rel":584},"https:\u002F\u002Fonlinelibrary.wiley.com\u002Fdoi\u002F10.1002\u002Fejsp.674",[236],"DOI: 10.1002\u002Fejsp.674",[33,587,245],{"href":294,"ariaLabel":295,"className":588,"dataFootnoteBackref":38},[244],[33,590,245,592],{"href":299,"ariaLabel":300,"className":591,"dataFootnoteBackref":38},[244],[30,593,59],{},[142,595,596,597,600,601,238,606,246,609],{"id":306},"Linardon, J., Cuijpers, P., Carlbring, P., Messer, M., & Fuller-Tyszkiewicz, M. (2019). The efficacy of app-supported smartphone interventions for mental health problems: A meta-analysis of randomized controlled trials. ",[87,598,599],{},"World Psychiatry",", 18(3), 325-336. ",[33,602,605],{"href":603,"rel":604},"https:\u002F\u002Fonlinelibrary.wiley.com\u002Fdoi\u002Ffull\u002F10.1002\u002Fwps.20673",[236],"DOI: 10.1002\u002Fwps.20673",[33,607,245],{"href":318,"ariaLabel":319,"className":608,"dataFootnoteBackref":38},[244],[33,610,245,614],{"href":611,"ariaLabel":612,"className":613,"dataFootnoteBackref":38},"#user-content-fnref-4-2","Back to reference 4-2",[244],[30,615,59],{},{"title":38,"searchDepth":367,"depth":367,"links":617},[618,619,620,621,622,623],{"id":398,"depth":367,"text":399},{"id":414,"depth":367,"text":415},{"id":433,"depth":367,"text":434},{"id":452,"depth":367,"text":453},{"id":468,"depth":367,"text":469},{"id":37,"depth":367,"text":220},"Self-monitoring is one of the best-evidenced techniques in behavioral science, yet most tracking apps fail in real-world use. What the research shows about that gap, and what it implies for design.",{},"\u002Fblog\u002Fdo-habit-trackers-actually-work",{"title":390,"description":624},"blog\u002Fdo-habit-trackers-actually-work",[386,630,631],"behavior","product","mukNUH-ItJCqD7FjGzQ9FztsFzIdjLIaWdDcYu9V3B8",{"id":634,"title":635,"author":7,"body":636,"date":774,"description":775,"extension":377,"meta":776,"navigation":379,"path":777,"seo":778,"stem":779,"tags":780,"__hash__":781},"blog\u002Fblog\u002Fnotes-vs-numbers-what-journaling-adds.md","Notes vs numbers: journaling vs statistics",{"type":9,"value":637,"toc":767},[638,641,645,658,675,679,685,692,696,703,706,710,713,718,726,730],[15,639,640],{},"Daynostic works mainly with numbers: it turns each day into structured data and searches for statistically defensible patterns. It also asks for a sentence or two of free writing each night, which sits outside everything the correlation engine can process. There are two reasons to ask for it, both supported by a research literature older than most tracking apps.",[22,642,644],{"id":643},"the-value-of-the-writing-itself","The value of the writing itself",[15,646,647,648,651,652,657],{},"In 1986, James Pennebaker and Sandra Beall ran the experiment that started the field. Participants wrote for a few minutes on four consecutive days, some about the most difficult experiences of their lives and others about trivial topics. The group that wrote about difficult experiences showed ",[145,649,650],{},"better long-term health outcomes"," in the months that followed.",[30,653,654],{},[33,655,40],{"href":35,"ariaDescribedBy":656,"dataFootnoteRef":38,"id":39},[37]," Putting experience into words, without sharing or analysing it, appeared to do something measurable on its own.",[15,659,660,661,664,665,668,669,674],{},"Decades of replication refined the claim to ",[87,662,663],{},"real but modest",". The most comprehensive meta-analysis, by Joanne Frattaroli, pooled ",[145,666,667],{},"146 randomized studies"," of this kind of expressive disclosure and found a small but statistically significant benefit to health and well-being, larger for people under high stress.",[30,670,671],{},[33,672,59],{"href":56,"ariaDescribedBy":673,"dataFootnoteRef":38,"id":58},[37]," The effect is small, but it holds independently of any dashboard, correlation or insight. The writing itself is the intervention.",[22,676,678],{"id":677},"the-context-a-number-cannot-hold","The context a number cannot hold",[15,680,681,682,684],{},"A mood rating of ",[145,683,59],{}," is clean and comparable, but it carries no reason. A 2 after a bad night's sleep, a 2 on the day a pet died, and a 2 during a stressful deadline are identical as numbers and entirely different as experiences. The reason goes into the note.",[15,686,687,688,691],{},"This also marks a hard technical boundary. ",[145,689,690],{},"Free text cannot enter the correlation engine."," The statistics require structured, comparable, numeric variables; a permutation test cannot run on \"felt off today, maybe the weather\". The two formats answer different questions. The numbers answer what is reliably associated with what. The notes record what was going on. Neither format can do the other's job.",[22,693,695],{"id":694},"where-ai-insights-fit","Where AI insights fit",[15,697,698,699,702],{},"Because the prose sits outside the statistics, it needs a different kind of reader, and that is the role for AI in a tool like this. The statistics handle everything they can test rigorously. A language model is pointed only at the part ",[33,700,701],{"href":196},"the pipeline cannot process",": reading the notes for themes, context and the qualitative thread that no correlation will find. That is the one job in the app where a language model does something the statistics cannot.",[15,704,705],{},"The diary's tight character limit follows the same logic. The limit is there to keep the entry to what mattered that day. A few deliberate sentences are more useful than a long entry, both for the writer and for anything that reads them later.",[22,707,709],{"id":708},"two-instruments","Two instruments",[15,711,712],{},"Numbers and notes are best treated as two instruments recording the same day. The number is comparable, testable and eventually correlatable across hundreds of entries. The sentence is personal and incomparable, and it holds the meaning a number leaves out. One can show that low days cluster on poor-sleep nights; the other records what those days were like.",[200,714,715],{},[15,716,717],{},"The number is for the statistics; the sentence is for the context. Together they record more of a day than either does alone, in the same half-minute before bed.",[15,719,720,721,725],{},"Neither should be oversold. Expressive writing is a small, genuine benefit, not a treatment, and a correlation is ",[33,722,724],{"href":723},"\u002Fblog\u002Fcorrelation-isnt-causation","a question, not a verdict",". Kept in their proper roles, the two formats describe a life more completely than numbers alone could.",[206,727,728],{},[15,729,210],{},[212,731,733,736],{"className":732,"dataFootnotes":38},[215],[22,734,220],{"className":735,"id":37},[219],[222,737,738,753],{},[142,739,740,741,744,745,238,750],{"id":226},"Pennebaker, J. W., & Beall, S. K. (1986). Confronting a traumatic event: Toward an understanding of inhibition and disease. ",[87,742,743],{},"Journal of Abnormal Psychology",", 95(3), 274-281. ",[33,746,749],{"href":747,"rel":748},"https:\u002F\u002Fpubmed.ncbi.nlm.nih.gov\u002F3745650\u002F",[236],"PubMed 3745650",[33,751,245],{"href":241,"ariaLabel":242,"className":752,"dataFootnoteBackref":38},[244],[142,754,755,756,758,759,238,764],{"id":256},"Frattaroli, J. (2006). Experimental disclosure and its moderators: A meta-analysis. ",[87,757,539],{},", 132(6), 823-865. ",[33,760,763],{"href":761,"rel":762},"https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FExperimental-disclosure-and-its-moderators:-a-Frattaroli\u002F4cda801628a1d3fcb126fd0ce1db22d720245017",[236],"Semantic Scholar",[33,765,245],{"href":269,"ariaLabel":270,"className":766,"dataFootnoteBackref":38},[244],{"title":38,"searchDepth":367,"depth":367,"links":768},[769,770,771,772,773],{"id":643,"depth":367,"text":644},{"id":677,"depth":367,"text":678},{"id":694,"depth":367,"text":695},{"id":708,"depth":367,"text":709},{"id":37,"depth":367,"text":220},"2026-06-12","Daynostic runs on numbers and journaling. The research on expressive writing, why free text cannot enter a correlation, and how notes and numbers play different roles.",{},"\u002Fblog\u002Fnotes-vs-numbers-what-journaling-adds",{"title":635,"description":775},"blog\u002Fnotes-vs-numbers-what-journaling-adds",[385,631],"SLXI_3J5Cn-hccSiXpSAedBXTIXBS61X6eJ2SUCnqpY",{"id":783,"title":784,"author":7,"body":785,"date":985,"description":986,"extension":377,"meta":987,"navigation":379,"path":988,"seo":989,"stem":990,"tags":991,"__hash__":992},"blog\u002Fblog\u002Frecall-bias-why-you-cant-trust-your-memory.md","Recall bias: why memory of last month is unreliable",{"type":9,"value":786,"toc":978},[787,792,795,799,820,846,850,855,858,869,873,889,892,896,899,902,907,911],[12,788,789],{},[15,790,791],{},"Memory does not store the past; it reconstructs it.",[15,793,794],{},"A quick sense of how last month went arrives almost instantly: \"stressful\", \"pretty good\", \"a blur\". It feels like recall, but it is mostly reconstruction, assembled on the spot from a few vivid moments and a general theory about the period, and it is wrong in specific, predictable ways. Understanding how it is wrong is the strongest argument for recording days as they happen.",[22,796,798],{"id":797},"the-peak-end-rule","The peak-end rule",[15,800,801,802,805,806,811,812,815,816,819],{},"The clearest demonstration comes from an uncomfortable study. Redelmeier and Kahneman tracked patients' pain ",[87,803,804],{},"during"," colonoscopies, moment by moment, then measured how bad they remembered the procedure afterward.",[30,807,808],{},[33,809,40],{"href":35,"ariaDescribedBy":810,"dataFootnoteRef":38,"id":39},[37]," Remembered unpleasantness barely depended on how long the procedure lasted or the total pain involved. It was predicted almost entirely by two moments: the ",[87,813,814],{},"worst"," (the peak) and the ",[87,817,818],{},"final"," (the end).",[15,821,822,823,826,827,832,833,836,837,840,841],{},"This is the ",[145,824,825],{},"peak-end rule",", shown earlier in a cleaner experiment where people held a hand in painful cold water and remembered discomfort tracked the average of the peak and the end while almost ignoring duration.",[30,828,829],{},[33,830,59],{"href":56,"ariaDescribedBy":831,"dataFootnoteRef":38,"id":58},[37]," A later randomized trial of 682 colonoscopy patients went further: ",[87,834,835],{},"adding"," extra time at the end, more total discomfort but milder than the peak, left patients remembering the longer procedure as ",[145,838,839],{},"less"," unpleasant, and made them more likely to return for a follow-up.",[30,842,843],{},[33,844,81],{"href":78,"ariaDescribedBy":845,"dataFootnoteRef":38,"id":80},[37],[22,847,849],{"id":848},"duration-neglect","Duration neglect",[15,851,852,854],{},[145,853,849],{},", the finding that the length of an experience barely registers in memory, applies well beyond medical procedures. A sense of \"a rough month\" is dominated by its worst day and by how it ended, not by the distribution of its forty-odd waking experiences. Three calm weeks and one difficult Thursday can be filed as \"a difficult month\", because the peak and the most recent moments dominate.",[15,856,857],{},"This makes memory a poor instrument for the question a life tracker is concerned with: how things have been. A remembered average is a selection of a few moments, chosen by biases that operate without being noticed.",[15,859,860,861,864,865,868],{},"A concrete case appeared earlier: ",[33,862,863],{"href":380},"in a study of 340,000 people",", Mondays were no worse than other weekdays in logged mood, yet 65% of people ",[87,866,867],{},"remembered"," Monday as their worst day. The belief did not match the record.",[22,870,872],{"id":871},"why-clinical-science-stopped-relying-on-recall","Why clinical science stopped relying on recall",[15,874,875,876,879,880,883,884],{},"Researchers met this problem decades ago and changed their methods. Instead of asking \"how was your week?\" at a clinic visit, which is exactly the recall-biased question, they adopted ",[145,877,878],{},"Ecological Momentary Assessment (EMA)",": capturing experience ",[87,881,882],{},"in the moment",", repeatedly, in everyday life. As Shiffman, Stone and Hufford put it in their review, the purpose of EMA is to minimize recall bias and maximize ecological validity, recording data before memory can rewrite it.",[30,885,886],{},[33,887,98],{"href":95,"ariaDescribedBy":888,"dataFootnoteRef":38,"id":97},[37],[15,890,891],{},"The principle extends past research: an accurate picture of how a stretch of time felt has to be measured close to the experience, not reconstructed from it later.",[22,893,895],{"id":894},"a-nightly-check-in-is-a-small-ema","A nightly check-in is a small EMA",[15,897,898],{},"A thirty-second evening log is a small version of the same method: a record of the day taken before the peak-end edit happens. Over a few months it produces something memory cannot, the actual distribution of days, with every ordinary Tuesday weighted as heavily as the dramatic ones.",[15,900,901],{},"A year-at-a-glance view serves the same purpose. A grid of every day, all present and equally sized, shows the calm weeks that would otherwise be forgotten and keeps one difficult day from standing for the whole month. On the question of how a period went, the record is more reliable than the memory of it.",[200,903,904],{},[15,905,906],{},"Logging a day while it is fresh records what memory will later overwrite with its worst and most recent moments.",[206,908,909],{},[15,910,210],{},[212,912,914,917],{"className":913,"dataFootnotes":38},[215],[22,915,220],{"className":916,"id":37},[219],[222,918,919,934,949,963],{},[142,920,921,922,925,926,238,931],{"id":226},"Redelmeier, D. A., & Kahneman, D. (1996). Patients' memories of painful medical treatments: Real-time and retrospective evaluations of two minimally invasive procedures. ",[87,923,924],{},"Pain",", 66(1), 3-8. ",[33,927,930],{"href":928,"rel":929},"https:\u002F\u002Fpubmed.ncbi.nlm.nih.gov\u002F8857625\u002F",[236],"PubMed 8857625",[33,932,245],{"href":241,"ariaLabel":242,"className":933,"dataFootnoteBackref":38},[244],[142,935,936,937,940,941,238,946],{"id":256},"Kahneman, D., Fredrickson, B. L., Schreiber, C. A., & Redelmeier, D. A. (1993). When more pain is preferred to less: Adding a better end. ",[87,938,939],{},"Psychological Science",", 4(6), 401-405. ",[33,942,945],{"href":943,"rel":944},"https:\u002F\u002Fdoi.org\u002F10.1111\u002Fj.1467-9280.1993.tb00589.x",[236],"DOI: 10.1111\u002Fj.1467-9280.1993.tb00589.x",[33,947,245],{"href":269,"ariaLabel":270,"className":948,"dataFootnoteBackref":38},[244],[142,950,951,952,954,955,238,960],{"id":281},"Redelmeier, D. A., Katz, J., & Kahneman, D. (2003). Memories of colonoscopy: A randomized trial. ",[87,953,924],{},", 104(1-2), 187-194. ",[33,956,959],{"href":957,"rel":958},"https:\u002F\u002Fpubmed.ncbi.nlm.nih.gov\u002F12855328\u002F",[236],"PubMed 12855328",[33,961,245],{"href":294,"ariaLabel":295,"className":962,"dataFootnoteBackref":38},[244],[142,964,965,966,969,970,238,975],{"id":306},"Shiffman, S., Stone, A. A., & Hufford, M. R. (2008). Ecological momentary assessment. ",[87,967,968],{},"Annual Review of Clinical Psychology",", 4, 1-32. ",[33,971,974],{"href":972,"rel":973},"https:\u002F\u002Fdoi.org\u002F10.1146\u002Fannurev.clinpsy.3.022806.091415",[236],"DOI: 10.1146\u002Fannurev.clinpsy.3.022806.091415",[33,976,245],{"href":318,"ariaLabel":319,"className":977,"dataFootnoteBackref":38},[244],{"title":38,"searchDepth":367,"depth":367,"links":979},[980,981,982,983,984],{"id":797,"depth":367,"text":798},{"id":848,"depth":367,"text":849},{"id":871,"depth":367,"text":872},{"id":894,"depth":367,"text":895},{"id":37,"depth":367,"text":220},"2026-06-11","Memory does not store the weeks that passed; it reconstructs them, badly. The peak-end rule, duration neglect, and why clinical science switched to in-the-moment logging.",{},"\u002Fblog\u002Frecall-bias-why-you-cant-trust-your-memory",{"title":784,"description":986},"blog\u002Frecall-bias-why-you-cant-trust-your-memory",[385,630],"CRwdlQ0rBaei-3XDd5jeLkoIliWunplrGssBzPvgeNI",{"id":994,"title":995,"author":7,"body":996,"date":1175,"description":1176,"extension":377,"meta":1177,"navigation":379,"path":1178,"seo":1179,"stem":1180,"tags":1181,"__hash__":1182},"blog\u002Fblog\u002Fcan-you-detect-your-own-seasonal-pattern.md","Can you detect your own seasonal pattern?",{"type":9,"value":997,"toc":1167},[998,1003,1006,1010,1023,1027,1040,1043,1047,1064,1067,1071,1074,1103,1107,1112,1115,1119],[12,999,1000],{},[15,1001,1002],{},"Winter makes me miserable.",[15,1004,1005],{},"One of the most common beliefs people hold about their own mood. It may be accurate, or it may be a cultural script attached to a few random bad weeks. The research on seasonal mood is unsettled, and that is the point of this article: whether a given person has a real seasonal pattern is a question that only that person's own data can answer.",[22,1007,1009],{"id":1008},"the-clinical-idea","The clinical idea",[15,1011,1012,1013,1016,1017,1022],{},"Seasonal mood swings received their clinical name in 1984, when Rosenthal and colleagues described ",[145,1014,1015],{},"Seasonal Affective Disorder",", a recurrent winter depression, and reported that extending the day with bright artificial light appeared to help.",[30,1018,1019],{},[33,1020,40],{"href":35,"ariaDescribedBy":1021,"dataFootnoteRef":38,"id":39},[37]," The proposed mechanism was straightforward: shorter winter days disrupt mood, and added light restores it. SAD entered the diagnostic manuals, light therapy became a standard recommendation, and the \"winter blues\" became common knowledge.",[22,1024,1026],{"id":1025},"the-population-evidence","The population evidence",[15,1028,1029,1030,1035,1036,1039],{},"Large studies later struggled to confirm it. In 2016, Traffanstedt, Mehta and LoBello analysed depression scores in a U.S. sample of more than 34,000 adults and tested them against season, latitude and sunlight exposure.",[30,1031,1032],{},[33,1033,59],{"href":56,"ariaDescribedBy":1034,"dataFootnoteRef":38,"id":58},[37]," Depression was ",[145,1037,1038],{},"unrelated"," to all three. The authors concluded that the data did not support seasonal variation as a general feature of major depression, which contradicts the assumption that winter reliably lowers mood across the population.",[15,1041,1042],{},"That leaves a recognised syndrome on one side and a large null result on the other. Both can be true at the same time.",[22,1044,1046],{"id":1045},"why-the-average-hides-the-individual","Why the average hides the individual",[15,1048,1049,1050,1053,1054,1057,1058,1063],{},"The largest careful study of weather and daily mood points to the reason. Denissen and colleagues found that the ",[145,1051,1052],{},"average"," effect of sunlight and day length on mood was small, while the variation ",[145,1055,1056],{},"between individuals"," was substantial, especially in how people responded to daylight.",[30,1059,1060],{},[33,1061,81],{"href":78,"ariaDescribedBy":1062,"dataFootnoteRef":38,"id":80},[37]," Some people are light-sensitive, many are not, and a few feel worse in summer.",[15,1065,1066],{},"Averaged across tens of thousands of people, those opposite responses cancel. A population study reports \"no seasonal effect\" because the seasonal minority is diluted by everyone who is unaffected, plus the people who respond the other way. For a trait that varies this much between people, the group mean says little about any individual.",[22,1068,1070],{"id":1069},"detecting-a-personal-signal","Detecting a personal signal",[15,1072,1073],{},"Finding a personal seasonal signal is one of the things daily tracking is suited for, and also one of the harder patterns to establish. Three reasons:",[139,1075,1076,1087,1093],{},[142,1077,1078,1081,1082,1086],{},[145,1079,1080],{},"A season is a slow cycle."," It spans a full year, so a single pass through winter provides one observation of the winter effect. As ",[33,1083,1085],{"href":1084},"\u002Fblog\u002Fhow-many-days-until-your-data-means-anything","the article on sample size"," explains, one cycle is far too little for confidence. Confirming a seasonal rhythm as such takes several years.",[142,1088,1089,1092],{},[145,1090,1091],{},"Daylight is a better variable than the calendar."," Treating daily sunlight, temperature and day length as continuous predictors, logged every day, turns a once-a-year signal into a daily one that is much easier to detect. Weather recorded automatically each day captures these variables.",[142,1094,1095,1098,1099,1102],{},[145,1096,1097],{},"Winter is a bundle of confounders."," Less daylight arrives together with less exercise, more time indoors, holiday stress, disrupted sleep and lower vitamin D. A simple \"December equals low mood\" result could reflect any of them, so ",[33,1100,1101],{"href":723},"the daylight signal has to be separated from the rest",".",[22,1104,1106],{"id":1105},"in-practice","In practice",[200,1108,1109],{},[15,1110,1111],{},"Neither the calendar nor the cultural script is a reliable guide, and a population null result does not rule out a personal pattern either. The test that works is a personal one: logging through the seasons, recording daylight and weather every day, and checking with proper statistics whether a suspected November dip repeats or was a few bad weeks that memory joined into a pattern.",[15,1113,1114],{},"A population study cannot answer the question for any single person. A year or two of consistent personal data can.",[206,1116,1117],{},[15,1118,210],{},[212,1120,1122,1125],{"className":1121,"dataFootnotes":38},[215],[22,1123,220],{"className":1124,"id":37},[219],[222,1126,1127,1142,1157],{},[142,1128,1129,1130,1133,1134,238,1139],{"id":226},"Rosenthal, N. E., Sack, D. A., Gillin, J. C., Lewy, A. J., Goodwin, F. K., Davenport, Y., Mueller, P. S., Newsome, D. A., & Wehr, T. A. (1984). Seasonal affective disorder: A description of the syndrome and preliminary findings with light therapy. ",[87,1131,1132],{},"Archives of General Psychiatry",", 41(1), 72-80. ",[33,1135,1138],{"href":1136,"rel":1137},"https:\u002F\u002Fdoi.org\u002F10.1001\u002Farchpsyc.1984.01790120076010",[236],"DOI: 10.1001\u002Farchpsyc.1984.01790120076010",[33,1140,245],{"href":241,"ariaLabel":242,"className":1141,"dataFootnoteBackref":38},[244],[142,1143,1144,1145,1148,1149,238,1154],{"id":256},"Traffanstedt, M. K., Mehta, S., & LoBello, S. G. (2016). Major depression with seasonal variation: Is it a valid construct? ",[87,1146,1147],{},"Clinical Psychological Science",", 4(5), 825-834. ",[33,1150,1153],{"href":1151,"rel":1152},"https:\u002F\u002Fjournals.sagepub.com\u002Fdoi\u002Fabs\u002F10.1177\u002F2167702615615867",[236],"Abstract (SAGE)",[33,1155,245],{"href":269,"ariaLabel":270,"className":1156,"dataFootnoteBackref":38},[244],[142,1158,227,1159,231,1161,238,1164],{"id":281},[87,1160,230],{},[33,1162,237],{"href":234,"rel":1163},[236],[33,1165,245],{"href":294,"ariaLabel":295,"className":1166,"dataFootnoteBackref":38},[244],{"title":38,"searchDepth":367,"depth":367,"links":1168},[1169,1170,1171,1172,1173,1174],{"id":1008,"depth":367,"text":1009},{"id":1025,"depth":367,"text":1026},{"id":1045,"depth":367,"text":1046},{"id":1069,"depth":367,"text":1070},{"id":1105,"depth":367,"text":1106},{"id":37,"depth":367,"text":220},"2026-06-09","Seasonal mood swings are real for some people and absent for most, and population research cannot say which applies to a given person. Why daylight, latitude and season cancel out on average, and why a year of personal data is the only way to find the signal.",{},"\u002Fblog\u002Fcan-you-detect-your-own-seasonal-pattern",{"title":995,"description":1176},"blog\u002Fcan-you-detect-your-own-seasonal-pattern",[385,384],"aTyeiwmgUNG_UxrxGzOn0lAWt43Kz0MaKMXy5RawlOQ",{"id":1184,"title":1185,"author":7,"body":1186,"date":1291,"description":1292,"extension":377,"meta":1293,"navigation":379,"path":1294,"seo":1295,"stem":1296,"tags":1297,"__hash__":1300},"blog\u002Fblog\u002Fyour-location-never-leaves-your-phone.md","Your location never leaves your phone",{"type":9,"value":1187,"toc":1286},[1188,1193,1196,1199,1203,1206,1234,1237,1241,1244,1247,1267,1271,1274,1281],[12,1189,1190],{},[15,1191,1192],{},"Daynostic can use weather data without ever learning where a person is.",[15,1194,1195],{},"Weather is one of the more interesting variables in life tracking. Temperature swings, pressure drops and long stretches of low light have been suspected for centuries of affecting mood and energy, and a daily mood log paired with daily weather data makes that testable at the individual level.",[15,1197,1198],{},"The obstacle is location. Knowing the weather requires knowing where someone is, and location is among the most sensitive signals a phone produces: a year of daily coordinates maps a home, a workplace, travel and routine. The aim was to obtain the weather data without ever assembling that map.",[22,1200,1202],{"id":1201},"the-lookup-happens-on-the-device","The lookup happens on the device",[15,1204,1205],{},"When Daynostic records weather for a day, the sequence is:",[222,1207,1208,1214,1225,1231],{},[142,1209,1210,1213],{},[145,1211,1212],{},"The device"," determines its own location, as any weather app does.",[142,1215,1216,1218,1219,1224],{},[145,1217,1212],{}," calls ",[33,1220,1223],{"href":1221,"rel":1222},"https:\u002F\u002Fopen-meteo.com",[236],"Open-Meteo",", an open weather API, directly. The request travels from the phone to the weather service and never reaches Daynostic's servers.",[142,1226,1227,1228,1102],{},"The response (temperature, humidity, atmospheric pressure, wind, precipitation) is attached to the day ",[145,1229,1230],{},"locally",[142,1232,1233],{},"With sync enabled, only those daily weather numbers reach the account. Coordinates are never part of the payload.",[15,1235,1236],{},"Daynostic's servers can therefore record that a given day was 18 °C with light rain, without knowing where on Earth that day took place. There is no location column in the database, because the data never arrives.",[22,1238,1240],{"id":1239},"trust-us-versus-we-cant","\"Trust us\" versus \"we can't\"",[15,1242,1243],{},"Most privacy policies are promises: location is collected, and the company commits to handling it carefully. That is a \"trust us\" design, and it depends on every future version of the company behaving well.",[15,1245,1246],{},"A \"we can't\" design is stronger. Data that is never transmitted cannot be misused, whatever anyone's intentions, and an architectural guarantee is much easier to verify than a stated one. The same approach appears elsewhere in Daynostic:",[139,1248,1249,1255,1261],{},[142,1250,1251,1254],{},[145,1252,1253],{},"Guest mode is fully local."," The app works without an account, and everything stays on the device until the user chooses to sync.",[142,1256,1257,1260],{},[145,1258,1259],{},"Data is exportable."," Logged data belongs to the person who recorded it and can be taken out at any time in an open format.",[142,1262,1263,1266],{},[145,1264,1265],{},"Deleting an account deletes its data."," No copies are kept for analytics.",[22,1268,1270],{"id":1269},"why-open-meteo","Why Open-Meteo",[15,1272,1273],{},"Open-Meteo is an open-source weather API that needs no API key or user account for the variables Daynostic uses. That is what makes the direct-from-device call possible: there is no secret credential that would force the request through a backend. It supplies the daily variables relevant to correlation analysis: temperature, humidity, pressure, wind and precipitation.",[15,1275,1276,1277,1280],{},"Because weather arrives automatically each day, the analysis receives a dense, objective data stream with no extra effort from the user. Once there is enough data, Daynostic can report, with the ",[33,1278,1279],{"href":196},"statistical backing described elsewhere",", whether something like pressure-related headaches is real for a given person. Until then it reports nothing.",[206,1282,1283],{},[15,1284,1285],{},"Daynostic's servers store daily weather values but never a location. The guarantee is architectural: the coordinates are not transmitted, so there is nothing to leak.",{"title":38,"searchDepth":367,"depth":367,"links":1287},[1288,1289,1290],{"id":1201,"depth":367,"text":1202},{"id":1239,"depth":367,"text":1240},{"id":1269,"depth":367,"text":1270},"2026-06-08","How Daynostic collects weather on the device, through Open-Meteo, so coordinates are never sent to its servers.",{},"\u002Fblog\u002Fyour-location-never-leaves-your-phone",{"title":1185,"description":1292},"blog\u002Fyour-location-never-leaves-your-phone",[1298,1299],"privacy","engineering","kFoFcSoyA34CCGnXduCtrpNUk1mucVUI9tqjFuzhRTM",{"id":1302,"title":1303,"author":7,"body":1304,"date":1464,"description":1465,"extension":377,"meta":1466,"navigation":379,"path":1467,"seo":1468,"stem":1469,"tags":1470,"__hash__":1471},"blog\u002Fblog\u002Fsleep-and-mood-what-averages-hide.md","Sleep and mood: what averages hide",{"type":9,"value":1305,"toc":1457},[1306,1311,1314,1318,1339,1346,1350,1362,1365,1367,1370,1380,1384,1387,1417,1420,1425,1429],[12,1307,1308],{},[15,1309,1310],{},"Bad sleep means a bad day.",[15,1312,1313],{},"This is one of the few folk-psychology claims the research broadly supports. The details are more interesting than the headline. The relationship has a direction, it depends more on perceived sleep than on measured sleep, and its strength varies a lot from one person to the next. That last point is the reason personal data can show something no study of strangers can.",[22,1315,1317],{"id":1316},"the-link-and-its-direction","The link, and its direction",[15,1319,1320,1321,1324,1325,1330,1331,1334,1335,1338],{},"The careful evidence comes from ",[87,1322,1323],{},"naturalistic"," studies, where people log their sleep and mood in daily life over many days rather than being measured once in a lab. A systematic review by Konjarski and colleagues gathered these prospective diary studies and found a consistent pattern.",[30,1326,1327],{},[33,1328,40],{"href":35,"ariaDescribedBy":1329,"dataFootnoteRef":38,"id":39},[37]," Day-to-day sleep and mood are linked, and the link is usually ",[145,1332,1333],{},"directional",": poor sleep predicts a worse ",[87,1336,1337],{},"next"," day more reliably than a poor mood predicts worse sleep that night.",[15,1340,1341,1342,1345],{},"Direction is what makes the pattern credible. When last night's sleep quality predicts the following day's mood, the cause precedes the effect, which satisfies the one ",[33,1343,1344],{"href":723},"Bradford Hill criterion that is strictly required"," and removes the same-day ambiguity that could otherwise run either way.",[22,1347,1349],{"id":1348},"perceived-sleep-vs-measured-sleep","Perceived sleep vs measured sleep",[15,1351,1352,1353,1356,1357],{},"A result that often surprises people is that in these studies ",[145,1354,1355],{},"subjective"," sleep, meaning how rested a person feels, tends to track next-day mood better than objective readings from a device. The same review noted that actigraphy-based sleep metrics did not show the clean coupling that self-reported sleep did.",[30,1358,1359],{},[33,1360,40],{"href":35,"ariaDescribedBy":1361,"dataFootnoteRef":38,"id":166},[37],[15,1363,1364],{},"This says something about how mood works, not about the accuracy of sleep trackers. The minutes spent in a given sleep stage predict next-day mood less well than a person's own sense of whether the night was restorative. A one-tap \"how was your sleep?\" rating captures that subjective sense, which makes it well suited to the question of whether sleep affects mood.",[22,1366,1046],{"id":1045},[15,1368,1369],{},"Reviews describe what holds on average across many people, but sleep-mood coupling is a trait with large individual differences. For some people one bad night reliably damages the next day. For others the effect is barely present. Averaging them gives a moderate effect that describes almost nobody precisely.",[15,1371,822,1372,1375,1376,1379],{},[33,1373,1374],{"href":380},"n-of-1 problem"," in miniature. A population study establishes that sleep-mood links ",[87,1377,1378],{},"exist","; it cannot say how strong the link is for a particular person. Only personal data, gathered over enough nights, can place someone on that spectrum. A strong personal coupling means protecting sleep is worth a lot of effort. A weak one means the bad days have some other cause.",[22,1381,1383],{"id":1382},"reading-it-in-personal-data","Reading it in personal data",[15,1385,1386],{},"Several practical points follow from the research:",[139,1388,1389,1395,1401,1411],{},[142,1390,1391,1394],{},[145,1392,1393],{},"The lag matters more than the same day."," Pairing last night's sleep with the next day's mood preserves the cause-before-effect ordering; same-day comparisons tangle the two.",[142,1396,1397,1400],{},[145,1398,1399],{},"Reverse causation runs the other way too."," An anxious or low evening can itself harm that night's sleep, so the arrow can point in both directions, which is something to test, not assume.",[142,1402,1403,1406,1407,1410],{},[145,1404,1405],{},"Detection takes months, not days."," A moderate effect ",[33,1408,1409],{"href":1084},"needs many observations","; a flat result after two weeks indicates insufficient data, not the absence of an effect.",[142,1412,1413,1416],{},[145,1414,1415],{},"Confounders move both variables."," Weekends, alcohol and stress shift sleep and mood together, so the analysis has to control for them before crediting sleep.",[15,1418,1419],{},"Daynostic's analysis is built for this kind of paired, lagged question with confounders controlled, and it leaves the question open until enough nights support an answer.",[200,1421,1422],{},[15,1423,1424],{},"Whether sleep governs mood cannot be settled in general. With consistent logging over a season, the data can show whether it holds for one person.",[206,1426,1427],{},[15,1428,210],{},[212,1430,1432,1435],{"className":1431,"dataFootnotes":38},[215],[22,1433,220],{"className":1434,"id":37},[219],[222,1436,1437],{},[142,1438,1439,1440,1443,1444,238,1449,246,1452],{"id":226},"Konjarski, M., Murray, G., Lee, V. V., & Jackson, M. L. (2018). Reciprocal relationships between daily sleep and mood: A systematic review of naturalistic prospective studies. ",[87,1441,1442],{},"Sleep Medicine Reviews",", 42, 47-58. ",[33,1445,1448],{"href":1446,"rel":1447},"https:\u002F\u002Fpubmed.ncbi.nlm.nih.gov\u002F30404728\u002F",[236],"PubMed 30404728",[33,1450,245],{"href":241,"ariaLabel":242,"className":1451,"dataFootnoteBackref":38},[244],[33,1453,245,1455],{"href":249,"ariaLabel":250,"className":1454,"dataFootnoteBackref":38},[244],[30,1456,59],{},{"title":38,"searchDepth":367,"depth":367,"links":1458},[1459,1460,1461,1462,1463],{"id":1316,"depth":367,"text":1317},{"id":1348,"depth":367,"text":1349},{"id":1045,"depth":367,"text":1046},{"id":1382,"depth":367,"text":1383},{"id":37,"depth":367,"text":220},"2026-06-07","The research on sleep and next-day mood is real, directional and strongly individual. Why population averages blur it, why perceived sleep predicts mood better than the clock, and what a season of personal logs can show.",{},"\u002Fblog\u002Fsleep-and-mood-what-averages-hide",{"title":1303,"description":1465},"blog\u002Fsleep-and-mood-what-averages-hide",[385,384],"_961xI2EfYx5WafNbb7sEqf_9KUINM6_90dSVVvbyZg",{"id":1473,"title":1474,"author":7,"body":1475,"date":1639,"description":1640,"extension":377,"meta":1641,"navigation":379,"path":1084,"seo":1642,"stem":1643,"tags":1644,"__hash__":1645},"blog\u002Fblog\u002Fhow-many-days-until-your-data-means-anything.md","How many days until the data means anything?",{"type":9,"value":1476,"toc":1632},[1477,1482,1485,1489,1500,1517,1547,1554,1558,1565,1572,1576,1579,1586,1590,1593,1600,1605,1609],[12,1478,1479],{},[15,1480,1481],{},"A week of data cannot reveal a reliable pattern.",[15,1483,1484],{},"A week of mood and sleep logs is enough for an app to compute a correlation, but not enough for any sound method to separate a real pattern from random noise. How many days personal data needs before it means something has a concrete, quantifiable answer, and it is useful to know it before trusting any number a tracker displays.",[22,1486,1488],{"id":1487},"statistical-power","Statistical power",[15,1490,1491,1492,1495,1496,1499],{},"Before running a study, researchers calculate ",[145,1493,1494],{},"statistical power",": the probability that, if a real effect exists, the study will have enough data to detect it. With too few participants, even a genuine effect stays buried in the noise, the study is ",[87,1497,1498],{},"underpowered",", and a null result means nothing.",[15,1501,1502,1503,1506,1507,1510,1511,1516],{},"Required sample size depends almost entirely on ",[145,1504,1505],{},"effect size",". Jacob Cohen's standard reference, ",[87,1508,1509],{},"A Power Primer",", tabulates it directly.",[30,1512,1513],{},[33,1514,40],{"href":35,"ariaDescribedBy":1515,"dataFootnoteRef":38,"id":39},[37]," Detecting a correlation with the conventional 80% power, at the usual significance threshold, takes roughly:",[139,1518,1519,1529,1538],{},[142,1520,1521,1524,1525,1528],{},[145,1522,1523],{},"~28 observations"," for a ",[87,1526,1527],{},"large"," effect (correlation near 0.5),",[142,1530,1531,1524,1534,1537],{},[145,1532,1533],{},"~85 observations",[87,1535,1536],{},"medium"," effect (near 0.3),",[142,1539,1540,1524,1543,1546],{},[145,1541,1542],{},"~780 observations",[87,1544,1545],{},"small"," effect (near 0.1).",[15,1548,1549,1550,1553],{},"A strong relationship can become detectable within a few weeks of daily logging. A weak one, and most effects in daily life are weak, can require ",[145,1551,1552],{},"two years"," of unbroken daily data before it is detectable at all. No algorithm gets around this.",[22,1555,1557],{"id":1556},"days-are-not-independent-observations","Days are not independent observations",[15,1559,1560,1561,1564],{},"It is tempting to treat each day as one clean observation, so that 780 observations means 780 days. It does not work that way, because of ",[145,1562,1563],{},"autocorrelation",": consecutive days are not independent. One day's mood resembles the previous day's, rough patches cluster into rough weeks, and sleep debt carries over. Weekly rhythms and slow trends add further structure.",[15,1566,1567,1568,1571],{},"When observations are correlated like this, the ",[87,1569,1570],{},"effective"," sample size is smaller than the raw day count, sometimes much smaller. A hundred logged days might carry the statistical weight of perhaps forty independent ones. The number of days required is therefore larger than Cohen's table suggests, and a feature that promises an insight after thirty days is not doing statistics.",[22,1573,1575],{"id":1574},"why-the-list-is-empty-at-first","Why the list is empty at first",[15,1577,1578],{},"Weak effects need hundreds of observations, and those observations are partly redundant. Together these mean that an empty insights list in the first weeks is the expected result of a correct analysis.",[15,1580,1581,1582,1585],{},"The threshold rises further when several variables are tracked at once. More variable pairs mean more chances for a coincidence to look significant, which is why serious analysis raises the threshold in proportion to the number of questions asked, a correction ",[33,1583,1584],{"href":380},"described separately",". With low power early on and many comparisons, the correct procedure is to wait until a pattern has enough data behind it to clear a deliberately high threshold. The strongest correlation in a small, noisy dataset is almost always noise, whatever an app chooses to do with it.",[22,1587,1589],{"id":1588},"more-days","More days",[15,1591,1592],{},"Power grows with every logged day. Each one moves small, real effects from undetectable toward visible and tightens the estimates on those already showing. There is no substitute for sample size; it can only be accumulated.",[15,1594,1595,1596,1599],{},"This is the logic behind an analysis ",[33,1597,1598],{"href":196},"held to a high statistical threshold",": the waiting period is the data accumulating until a result can be supported.",[200,1601,1602],{},[15,1603,1604],{},"A pattern from a handful of logs is almost always noise. Detecting a weak effect can take hundreds of observations, so an empty insights list in the first weeks is the expected result.",[206,1606,1607],{},[15,1608,210],{},[212,1610,1612,1615],{"className":1611,"dataFootnotes":38},[215],[22,1613,220],{"className":1614,"id":37},[219],[222,1616,1617],{},[142,1618,1619,1620,1622,1623,1628,1629],{"id":226},"Cohen, J. (1992). A power primer. ",[87,1621,539],{},", 112(1), 155-159. ",[33,1624,1627],{"href":1625,"rel":1626},"https:\u002F\u002Fdoi.org\u002F10.1037\u002F0033-2909.112.1.155",[236],"DOI: 10.1037\u002F0033-2909.112.1.155",". The cited sample sizes are for a product-moment correlation, two-tailed, α = .05, power = .80. ",[33,1630,245],{"href":241,"ariaLabel":242,"className":1631,"dataFootnoteBackref":38},[244],{"title":38,"searchDepth":367,"depth":367,"links":1633},[1634,1635,1636,1637,1638],{"id":1487,"depth":367,"text":1488},{"id":1556,"depth":367,"text":1557},{"id":1574,"depth":367,"text":1575},{"id":1588,"depth":367,"text":1589},{"id":37,"depth":367,"text":220},"2026-06-05","Why a pattern from a handful of logs is noise: what statistical power and sample size mean, how many observations are needed to detect an effect, and why an empty insights list at the start is the correct result.",{},{"title":1474,"description":1640},"blog\u002Fhow-many-days-until-your-data-means-anything",[384,385],"8U6SAvFHM3D2J10luPoEPo7LtzuGaiVzpuKRRyZsL_I",{"id":1647,"title":1648,"author":7,"body":1649,"date":1881,"description":1882,"extension":377,"meta":1883,"navigation":379,"path":723,"seo":1884,"stem":1885,"tags":1886,"__hash__":1887},"blog\u002Fblog\u002Fcorrelation-isnt-causation.md","Correlation isn't causation",{"type":9,"value":1650,"toc":1873},[1651,1656,1659,1663,1666,1703,1706,1710,1713,1716,1726,1730,1743,1783,1786,1790,1793,1817,1820,1824,1831,1836,1840],[12,1652,1653],{},[15,1654,1655],{},"Correlation isn't causation.",[15,1657,1658],{},"Probably the most quoted sentence in statistics, and a true one. The thinking often stops there, and the conclusion drawn is that correlations are worthless. That is wrong too. A correlation is not proof of cause, but it is evidence. This article covers what a correlation in personal data can indicate, what it cannot, and how science judges the difference.",[22,1660,1662],{"id":1661},"four-possible-explanations","Four possible explanations",[15,1664,1665],{},"Suppose the data show that mood tends to be higher on days with exercise. That single pattern is consistent with four quite different explanations:",[222,1667,1668,1674,1684,1694],{},[142,1669,1670,1673],{},[145,1671,1672],{},"Exercise improves mood",", the hoped-for reading.",[142,1675,1676,1679,1680,1683],{},[145,1677,1678],{},"Good mood prompts exercise",", which is ",[87,1681,1682],{},"reverse causation",": the run happens because the good mood is already there.",[142,1685,1686,1689,1690,1693],{},[145,1687,1688],{},"A third factor drives both",", a ",[87,1691,1692],{},"confounder",": on well-rested days both mood and the urge to exercise rise, and sleep is doing the real work.",[142,1695,1696,1699,1700,1102],{},[145,1697,1698],{},"Coincidence",": with enough variables, some pairs align by chance, a problem ",[33,1701,1702],{"href":380},"covered separately",[15,1704,1705],{},"A raw correlation cannot, on its own, distinguish between the four. It shows that two things move together and says nothing yet about why. That ambiguity is what the phrase means.",[22,1707,1709],{"id":1708},"confounders","Confounders",[15,1711,1712],{},"The textbook illustration: across a year, ice-cream sales correlate with drownings. Ice cream does not drown anyone; summer heat drives both. The confounder, hot weather, produces an association between two variables that have no direct link.",[15,1714,1715],{},"Daily life is full of the same structure. A correlation between coffee and productive days might reflect demanding workdays that raise both the coffee intake and the focus. \"Alcohol correlates with poor sleep\" is probably real, while \"social evenings correlate with poor sleep\" may be the same nights under a different name. A confounder is not visible in the pair being compared. It sits behind both variables and moves both, and one of them gets the credit.",[15,1717,193,1718,1720,1721],{},[87,1719,89],{}," project shows how readily numbers align by accident, pairing series such as cheese consumption and deaths by bedsheet entanglement: tight correlations, no meaning.",[30,1722,1723],{},[33,1724,40],{"href":35,"ariaDescribedBy":1725,"dataFootnoteRef":38,"id":39},[37],[22,1727,1729],{"id":1728},"the-bradford-hill-viewpoints","The Bradford Hill viewpoints",[15,1731,1732,1733,1736,1737,1742],{},"If a correlation is ambiguous, the question becomes how any field, such as medicine establishing that smoking causes cancer, ever moves from association to cause. In 1965 the epidemiologist Austin Bradford Hill set out the framework still in use: nine ",[87,1734,1735],{},"viewpoints"," for weighing whether an association is likely to be causal.",[30,1738,1739],{},[33,1740,59],{"href":56,"ariaDescribedBy":1741,"dataFootnoteRef":38,"id":58},[37]," They are meant to be weighed together rather than ticked off one by one. Several are directly useful for personal data:",[139,1744,1745,1759,1765,1771,1777],{},[142,1746,1747,1750,1751,1754,1755,1758],{},[145,1748,1749],{},"Temporality",": the cause has to precede the effect. This is the only strictly necessary viewpoint, and it is why pairing ",[87,1752,1753],{},"next-day"," mood with ",[87,1756,1757],{},"last night's"," sleep carries more weight than a same-day correlation. It fixes the direction and helps exclude reverse causation.",[142,1760,1761,1764],{},[145,1762,1763],{},"Strength",": a larger, clearer association is harder to dismiss as chance than a faint one.",[142,1766,1767,1770],{},[145,1768,1769],{},"Biological gradient",": a dose-response relationship, where more of the cause brings more of the effect, is strong evidence. Two coffees disrupting sleep more than one is more convincing than a yes-or-no association.",[142,1772,1773,1776],{},[145,1774,1775],{},"Consistency",": a pattern that recurs across time is more credible than one confined to a single month.",[142,1778,1779,1782],{},[145,1780,1781],{},"Plausibility",": a sensible mechanism strengthens the case, and its absence weakens it.",[15,1784,1785],{},"No single viewpoint is sufficient. Together they turn \"these two numbers move together\" into \"this is probably real, and probably in this direction.\"",[22,1787,1789],{"id":1788},"what-a-personal-correlation-is-good-for","What a personal correlation is good for",[15,1791,1792],{},"A correlation is useful in several ways, as long as it is treated as a question and not as an answer. It can:",[139,1794,1795,1801,1807],{},[142,1796,1797,1800],{},[145,1798,1799],{},"Generate hypotheses"," worth taking seriously, by pointing at a pattern in the data instead of a hunch.",[142,1802,1803,1806],{},[145,1804,1805],{},"Justify a deliberate test."," A suspected link between coffee and poor sleep can be checked with a small personal experiment, an n-of-1 trial that varies the coffee on purpose and observes the result. That is the route from correlation toward cause.",[142,1808,1809,1812,1813,1816],{},[145,1810,1811],{},"Rule things out."," A ",[87,1814,1815],{},"flat"," correlation over hundreds of days is informative in its own right: the feared or hoped-for link is not strong.",[15,1818,1819],{},"What a correlation cannot do is support a confident \"X causes Y\" from observation alone, which is why a tracker should not present one as an established cause.",[22,1821,1823],{"id":1822},"how-daynostic-handles-the-ambiguity","How Daynostic handles the ambiguity",[15,1825,1826,1827,1830],{},"Daynostic reports associations, attaches ",[33,1828,1829],{"href":196},"confidence tiers and a directional, next-day framing",", and does not present a correlation as a proven cause. A flagged result marks a strong, consistent, next-day link between two variables. It is a starting point for deliberate testing and stops short of any claim about what causes what.",[200,1832,1833],{},[15,1834,1835],{},"A correlation is best read as a question. It can point to a pattern worth attention, justify a deliberate test, or rule a weak link out. A confident statement about cause takes more than observation.",[206,1837,1838],{},[15,1839,210],{},[212,1841,1843,1846],{"className":1842,"dataFootnotes":38},[215],[22,1844,220],{"className":1845,"id":37},[219],[222,1847,1848,1858],{},[142,1849,307,1850,246,1852,238,1855],{"id":226},[87,1851,310],{},[33,1853,315],{"href":313,"rel":1854},[236],[33,1856,245],{"href":241,"ariaLabel":242,"className":1857,"dataFootnoteBackref":38},[244],[142,1859,1860,1861,1864,1865,238,1870],{"id":256},"Hill, A. B. (1965). The environment and disease: Association or causation? ",[87,1862,1863],{},"Proceedings of the Royal Society of Medicine",", 58(5), 295-300. ",[33,1866,1869],{"href":1867,"rel":1868},"https:\u002F\u002Fjournals.sagepub.com\u002Fdoi\u002F10.1177\u002F003591576505800503",[236],"Full text (SAGE)",[33,1871,245],{"href":269,"ariaLabel":270,"className":1872,"dataFootnoteBackref":38},[244],{"title":38,"searchDepth":367,"depth":367,"links":1874},[1875,1876,1877,1878,1879,1880],{"id":1661,"depth":367,"text":1662},{"id":1708,"depth":367,"text":1709},{"id":1728,"depth":367,"text":1729},{"id":1788,"depth":367,"text":1789},{"id":1822,"depth":367,"text":1823},{"id":37,"depth":367,"text":220},"2026-06-03","What a correlation in personal data can and cannot indicate: reverse causation, hidden confounders, and the Bradford Hill viewpoints science uses to judge when an association is worth believing.",{},{"title":1648,"description":1882},"blog\u002Fcorrelation-isnt-causation",[384,385],"th2oMLy8p7MgMp_B1EO7N4KtMWoBAkWo9FU0xJDRph0",{"id":1889,"title":1890,"author":7,"body":1891,"date":2036,"description":2037,"extension":377,"meta":2038,"navigation":379,"path":196,"seo":2039,"stem":2040,"tags":2041,"__hash__":2042},"blog\u002Fblog\u002Fhow-daynostic-decides-an-insight-is-real.md","How Daynostic decides an insight is real",{"type":9,"value":1892,"toc":2027},[1893,1896,1900,1909,1912,1916,1919,1928,1932,1935,1944,1948,1951,1955,1958,1969,1972,1976,1979,1983],[15,1894,1895],{},"Telling a real pattern apart from a coincidence is a statistics problem, and not a simple one. Two variables that drift together on a chart, or a handful of days that happen to line up, are not enough to call something an insight. Daynostic does not work by plotting two series, finding where they cross, and declaring a discovery. Behind every insight is an analysis pipeline assembled from established statistical methods, the kind used in scientific research and taught in graduate statistics courses. This article describes what that pipeline does to a set of daily check-ins before anything reaches the screen.",[22,1897,1899],{"id":1898},"rank-correlations","Rank correlations",[15,1901,1902,1903,1908],{},"Daynostic compares pairs of tracked variables, such as sleep quality and next-day mood, using a Spearman rank correlation.",[30,1904,1905],{},[33,1906,40],{"href":35,"ariaDescribedBy":1907,"dataFootnoteRef":38,"id":39},[37]," A Spearman correlation operates on the order of the values, not their raw magnitudes: it measures whether the days that rank highest on one variable tend to rank highest on the other.",[15,1910,1911],{},"The reason for working with ranks is that daily life data is noisy and poorly behaved. A single extreme day, food poisoning or a 2 a.m. fire alarm, can drag an ordinary average well off course while having almost no effect on a ranking. Rank-based methods also avoid treating a 1 to 5 mood scale as a precise measurement instrument. It is not one, and analysing ordinal ratings as if they were exact quantities is a common source of error in this kind of work.",[22,1913,1915],{"id":1914},"permutation-testing","Permutation testing",[15,1917,1918],{},"A correlation coefficient on its own says little. A coefficient of 0.4 across ten days carries almost no weight; the same value across several hundred days is a different matter. The textbook approach derives a p-value from a formula, but those formulas assume the data follow a particular distribution, which life data rarely does.",[15,1920,1921,1922,1927],{},"Daynostic uses a permutation test instead.",[30,1923,1924],{},[33,1925,59],{"href":56,"ariaDescribedBy":1926,"dataFootnoteRef":38,"id":58},[37]," It takes the actual data, shuffles the day labels thousands of times, and measures how often a correlation at least as strong as the observed one appears by chance in the reshuffled history. When random shuffles routinely produce a correlation as strong as the real one, the pattern cannot be separated from chance, and it is discarded. The more data behind a correlation, the harder it becomes for a shuffled version to match it by accident.",[22,1929,1931],{"id":1930},"correcting-for-multiple-comparisons","Correcting for multiple comparisons",[15,1933,1934],{},"The number of variable pairs grows quickly as more things are tracked. Testing many pairs raises the probability that at least one will appear significant by accident, a well understood issue in any field that runs large numbers of statistical tests at once.",[15,1936,1937,1938,1943],{},"Daynostic applies the Benjamini-Hochberg false-discovery-rate correction",[30,1939,1940],{},[33,1941,81],{"href":78,"ariaDescribedBy":1942,"dataFootnoteRef":38,"id":80},[37],", the same procedure used in genomics and clinical research when thousands of hypotheses are evaluated at the same time. The effect is that the more comparisons are made against the data, the higher the threshold each individual result has to clear.",[22,1945,1947],{"id":1946},"the-weekend-problem","The weekend problem",[15,1949,1950],{},"Saturdays differ from Tuesdays. People tend to sleep longer, exercise more and report better mood on weekends. Left uncorrected, this produces apparent correlations between almost everything and almost everything else, when the common cause is the structure of the week. Daynostic controls for day-of-week effects before judging a correlation, so a weekly rhythm is not mistaken for a discovery about sleep or exercise.",[22,1952,1954],{"id":1953},"confidence-tiers","Confidence tiers",[15,1956,1957],{},"Each association that passes the pipeline is labelled with one of three confidence tiers:",[139,1959,1960,1963,1966],{},[142,1961,1962],{},"Strong: passed every test with a clear margin, on a large number of days.",[142,1964,1965],{},"Moderate: passed the tests, but on fewer days or with a weaker effect.",[142,1967,1968],{},"Preliminary: an early result close to the threshold, which further data will either confirm or drop.",[15,1970,1971],{},"When no association passes, none is shown. A result appears once the statistics support it.",[22,1973,1975],{"id":1974},"putting-it-together","Putting it together",[15,1977,1978],{},"Each of these methods is standard in statistics. The work lies in combining them into a pipeline that runs automatically on a single person's data and holds to the same standards a researcher would apply by hand. An insight on the screen represents an association that has been ranked, tested against its own shuffled history, corrected for the number of comparisons made, and adjusted for the structure of the week. When Daynostic reports that sleep quality is associated with next-day mood, that statement has passed the same kind of scrutiny used to evaluate findings in published research.",[206,1980,1981],{},[15,1982,210],{},[212,1984,1986,1989],{"className":1985,"dataFootnotes":38},[215],[22,1987,220],{"className":1988,"id":37},[219],[222,1990,1991,2006,2016],{},[142,1992,1993,1994,1997,1998,238,2003],{"id":226},"Spearman, C. (1904). The proof and measurement of association between two things. ",[87,1995,1996],{},"The American Journal of Psychology",", 15(1), 72-101. ",[33,1999,2002],{"href":2000,"rel":2001},"https:\u002F\u002Fwww.jstor.org\u002Fstable\u002F1412159",[236],"JSTOR 1412159",[33,2004,245],{"href":241,"ariaLabel":242,"className":2005,"dataFootnoteBackref":38},[244],[142,2007,2008,2009,2012,2013],{"id":256},"The permutation (randomization) test originates with Fisher, R. A. (1935). ",[87,2010,2011],{},"The Design of Experiments."," Edinburgh: Oliver & Boyd. ",[33,2014,245],{"href":269,"ariaLabel":270,"className":2015,"dataFootnoteBackref":38},[244],[142,2017,2018,2019,328,2021,238,2024],{"id":281},"Benjamini, Y., & Hochberg, Y. (1995). Controlling the false discovery rate: a practical and powerful approach to multiple testing. ",[87,2020,327],{},[33,2022,333],{"href":331,"rel":2023},[236],[33,2025,245],{"href":294,"ariaLabel":295,"className":2026,"dataFootnoteBackref":38},[244],{"title":38,"searchDepth":367,"depth":367,"links":2028},[2029,2030,2031,2032,2033,2034,2035],{"id":1898,"depth":367,"text":1899},{"id":1914,"depth":367,"text":1915},{"id":1930,"depth":367,"text":1931},{"id":1946,"depth":367,"text":1947},{"id":1953,"depth":367,"text":1954},{"id":1974,"depth":367,"text":1975},{"id":37,"depth":367,"text":220},"2026-06-01","The statistical methods behind Daynostic's insights: Spearman rank correlations, permutation testing, false-discovery-rate correction, and controlling for day-of-week effects.",{},{"title":1890,"description":2037},"blog\u002Fhow-daynostic-decides-an-insight-is-real",[384,631],"C8kxuKQQ4mD-mg0yBmkvLFMLRiLgz9AM2I5K2_M5Hno",1794156286062]