If you’ve been collecting data on yourself, it’s now moderately easy to let an LLM write code that’ll tease out causal relationships between those variables, e.g. with a time series causal inference library like tigramite.
I’ve done that with some of my data, and surprisingly it both fails to back out results from RCTs I’ve run, and also doesn’t find many (to me) interesting causal relationships.
The most surprising one is that creatine influences creativity. Meditating more doesn’t influence anything (e.g. has no influence on mood or productivity), as far as could be determined. More mindful meditations cause happiness.
Because of the resampling/interpolation necessary, everything has a large autocorrelation.
Mainly posting this here so that people know it’s now easy to do this, and shill tigramite. I hope I can figure out a way of dealing with rare interventions (e.g. MDMA, CBD, ketamine), and incorporate more collected data. Code here, written by Claude 3.5.1 Sonnet.
Interesting. Do you have any recommendations on how to do this most effectively? At the moment I’m
using R to analyse the data/create some basic plots (will look into tigramite),
entering the data in google sheets,
occasionally doing a (blinded) RCT where I’m randomizing my dose of stimulants,
numerically tracking (guessing) mood and productivity,
and some things I mark as completed or not completed every day e.g. exercise, getting up early, remembering to floss
Questions I’d have:
Is google sheets good for something like this or are there better programs?
Any advice on blinding oneself or on RCTs in general or what variables one should track?
Anything I’m (probably) doing wrong?
Thanks in advance if you have any advice to offer (I already looked at your Caffeine RCT just wondering if you have any new insights or general advice on collecting data on oneself).
As for data collection, I’m probably currently less efficient than I could be. The bets guide on how to collect data is imho Passive measures for lazy self-experimenters (troof, 2022), I’d add that wearables like FitBit allow for data exporting (thanks GDPR!). I’ve written a bit about how I collect data here, which involves a haphazard combination of dmenu pop-ups, smartphone apps, manually invoked scripts and spreadsheets converted to CSV.
I’ve tried to err on the side of things that can automatically be collected, for anything that needs to be manually entered Google sheets is probably fine (though I don’t use it because I like to stay without internet most of the time).
As for blinding in RCTs[1], my process involves numbered envelopes containing both a pill and small piece of paper with a ‘P’ (placebo) or ‘I’ (intervention) written on it. Pills can be cut and put into pill capsules, sugar looks like a fine placebo.
I don’t have any great insight for what variables to track. I think from starting with the causal analysis I’ve updated towards tracking more “objective” measures (heart rate, sleeping times), and more things I can intervene on (though those usually have to be tracked manually).
If you’ve been collecting data on yourself, it’s now moderately easy to let an LLM write code that’ll tease out causal relationships between those variables, e.g. with a time series causal inference library like tigramite.
I’ve done that with some of my data, and surprisingly it both fails to back out results from RCTs I’ve run, and also doesn’t find many (to me) interesting causal relationships.
The most surprising one is that creatine influences creativity. Meditating more doesn’t influence anything (e.g. has no influence on mood or productivity), as far as could be determined. More mindful meditations cause happiness.
Because of the resampling/interpolation necessary, everything has a large autocorrelation.
Mainly posting this here so that people know it’s now easy to do this, and shill tigramite. I hope I can figure out a way of dealing with rare interventions (e.g. MDMA, CBD, ketamine), and incorporate more collected data. Code here, written by Claude 3.5.1 Sonnet.
Interesting. Do you have any recommendations on how to do this most effectively? At the moment I’m
using R to analyse the data/create some basic plots (will look into tigramite),
entering the data in google sheets,
occasionally doing a (blinded) RCT where I’m randomizing my dose of stimulants,
numerically tracking (guessing) mood and productivity,
and some things I mark as completed or not completed every day e.g. exercise, getting up early, remembering to floss
Questions I’d have:
Is google sheets good for something like this or are there better programs?
Any advice on blinding oneself or on RCTs in general or what variables one should track?
Anything I’m (probably) doing wrong?
Thanks in advance if you have any advice to offer (I already looked at your Caffeine RCT just wondering if you have any new insights or general advice on collecting data on oneself).
As for data collection, I’m probably currently less efficient than I could be. The bets guide on how to collect data is imho Passive measures for lazy self-experimenters (troof, 2022), I’d add that wearables like FitBit allow for data exporting (thanks GDPR!). I’ve written a bit about how I collect data here, which involves a haphazard combination of dmenu pop-ups, smartphone apps, manually invoked scripts and spreadsheets converted to CSV.
I’ve tried to err on the side of things that can automatically be collected, for anything that needs to be manually entered Google sheets is probably fine (though I don’t use it because I like to stay without internet most of the time).
As for blinding in RCTs[1], my process involves numbered envelopes containing both a pill and small piece of paper with a ‘P’ (placebo) or ‘I’ (intervention) written on it. Pills can be cut and put into pill capsules, sugar looks like a fine placebo.
I don’t have any great insight for what variables to track. I think from starting with the causal analysis I’ve updated towards tracking more “objective” measures (heart rate, sleeping times), and more things I can intervene on (though those usually have to be tracked manually).
Hope this helps :-)
I don’t think anyone has written up in detail how to do these! I should do that.