1,683
edits
Changes
Talk:Finding relations between variables in time series (edit)
Revision as of 22:32, 18 November 2023
, 22:32, 18 November 2023→another aggregator
https://www.lesswrong.com/posts/9kNxhKWvixtKW5anS/you-are-not-measuring-what-you-think-you-are-measuring 2 rules- takeways You are not measuring what you think you are measuring but enough data sources and types of things you measure you may find out what that is.
https://www.lesswrong.com/posts/9kNxhKWvixtKW5anS/you-are-not-measuring-what-you-think-you-are-measuring 2 rules- takeways You are not measuring what you think you are measuring but enough data sources and types of things you measure you may find out what that is.
https://gitlab.unige.ch/qol IMPORTANT
also eric j daza 's papers.
== funny on this stat analysis ==
== funny on this stat analysis ==
https://xkcd.com/2560/
https://xkcd.com/2560/
== Quick write I made here for later. ==
Collection is really just a matter of finding the right devices and taking the time to use them. Analysis outside of immediate obvious effect can become difficult. If the effect is subtle and drowned in other effects, or hard to measure. If the intervention is not something user can easily or wants to reproduce. If the effect take long time to build up, or is shifted in time from intervention. If the successful effect only happens under several conditions or several interventions together. If the spray and pray approach is dangerous. If the spray and pray approach only hits gold once in a while. Multiple comparison problem (see wikipedia). If user is bad at keeping records. There are probably more. There are many many apps that just do correlation and none that do anything more. Here is a list of both problems and apps.
== new section : single variable validity ==
how to prove that what you are measuring really is what you are trying to measure. aka construct validation
quick way; compare to a scientifically validated standard.
also consider en.wikipedia.org/wiki/Convergent_validity many tests all agree more or less and "divergent validity" they do not correlate with things that they should not
en.wikipedia.org/wiki/Nomological_network several constructs and their relationships to each other such as ageing causes memory loss
== some more thoughts ==
stats.stackexchange.com/questions/264225/finding-brief-repeated-patterns-in-a-time-series the question is not answered but in the side bar similar questions are well answered. Copulas may be important? The difference between time series and non-time series seems to be that with time series the patterns are cyclical, not a specific type of pattern/shape that happens every so often. Health tracking data seems to need both. I imagine blood sugar spike after meals but meals are not eaten at constant time. Long term trend and smoothing is covered in some Arima like models. Also changepoint which is like trend but in shorter time. And outliers? THat is like changepoint. results in a pretty visualization illustration of a single time series. If any of the apps were serious this would appear there.
Maybe multiple moving averages www.investopedia.com/terms/g/guppy-multiple-moving-average.asp as which kernel width fits best. Maybe ecg decomposition with DWT and ICA
This all called: single subject longitudinal analysis . how about Temporal Dynamics?
== research chat suggests ==
look at the extremes of the predictor variables; like three day long terrible sleep and three day long good sleep and compare cognitive ability
== another aggregator ==
www.opencures.org
apps.apple.com/us/app/this-that/id1660363624
== Series analysis is not it ==
Old concept about constantly doing old school statistical testing to know when there is enough data to stop something like a clinical trial.