Modeling correlates of personal happiness / anxiety

One of my reasons for self-quantification is to find strategies to increase days characterized by a positive mood and to reduce days characterized by agitation, stress, or anxiety. I have ideas about what I can focus on (for example, last month over three-quarters of my most positive daily experiences involved connecting with people, exercise, and appreciating nature; my most negative daily experiences were tagged as either anxiety or physical discomfort), but I want to design some data collection that I can use to actually model predictors (that are within my control) of my personal happiness / unhappiness.

For the model response variables I think I would develop rating scales for the emotional fluctuations I want to learn more about (e.g., anxiety, happiness). For the predictors, I expect that things like exercise, sleep, diet, and social connection are significant drivers of my mood. I already track my exercise, and I think the others could be fairly easy to monitor except for diet which I am somewhat resistant to (mostly the effort involved in accurately recording what I eat). Other ideas: some measure of flow, some measure of goal completion.

Has anywhere here done this or have any ideas about potentially important explanatory variables I could incorporate into my data collection?

2 Likes

Hi Erika,
Very interesting. I am six months into a similar self-tracking project on this topic. My analysis covers a good one and half years of data quantifying health observations, vitals, medicines, stool scale, activity and many more areas for myself and my family.

I am trying to slowly chip away at the complexity in this area. When I started the project, I had a set of assumptions regarding predictors, which I now realize are either not correct or a very small fraction of what will ultimately prove to be the complete set.

I am confident that a project like this requires self-tracking data covering multiple dimensions of well-being - from physical health, to economic/financial, to social/environment, intellectual/psychological and emotional/spiritual.

When analyzing a recent episode of stress/anxiety, I concluded that the predictors (whatever they may be) accumulate day-over-day, until a tipping point that then catalyzes an episode of stress/anxiety. In this scenario, I agree with your findings that some predictors like social interactions, sleep etc. help to reduce this accumulation. An accurate model in this area would identify the balance required to avoid the tipping point.

Sergio

2 Likes

Sounds like an interesting project! Please keep us posted on how it goes.

Myself and a few others have been working on large(ish) research project where participants are using a web application to track mood and up to 14 factors (e.g., sleep, food, exercise, social activity). From that, we’ve been modeling predictors of mood and then providing recommendations to participants about what might help them. An obvious point, but in comparing different sources of data for the modeling, we’ve found that the mere presence of a factor doesn’t give too much predictive power (e.g., just saying you were with person X). Instead, it seems that including your own appraisal rating really helps with modeling and especially if it’s a factor that has a big magnitude but changes in valence. (We included in the app an option to rate factors from -3 to +3 in terms of neg or pos impact.) Again obvious point but in the early stages of your tracking you may find it useful to include the appraisal component and then cut it out later. The modeling we did was from daily tracking for 3-4 weeks.

If you’re interested, I’m happy to talk by email (hollis@ucsc.edu) and can possibly share some other findings from our research which may or may not be helpful for your tracking. (For example, range of factors tracked and which ones seemed to most impact daily mood.)

2 Likes

I’ve thought about doing this quite a bit. If you make a model that takes a bunch of variables (sleep quality, social quality, exercise metrics, etc.) and predicts a single one (mood), you can use statistics to actually pinpoint which ones are the most predictive. Here’s a link on how to do it: http://blog.minitab.com/blog/adventures-in-statistics-2/how-to-identify-the-most-important-predictor-variables-in-regression-models

I want to do this for myself soon, and I really hope to see your results and wish you the best! :slight_smile:

I have done something similar but trying to predict what i call health quality index (HQI).

My takeaways from the study (which is still ongoing) is the following:

  • tracking needs to be quick and easy to do otherwise compliance will be low
  • I would predict several sets of days out. for example, i did 3 and 7. This allows you to see which affects happen over a short term and long term window
  • food is hard to track (i have found)
  • I started with simple linear prediction and on variable to vary. Once I got that working, then i moved up to many variables.

You can see my writeup here: https://github.com/isaacgerg/health_analysis/blob/master/manuscript/Managing_IBS_Through_Lightweight_Tracking.pdf

I found these videos to be a good refresher about the topic of regression: https://www.youtube.com/watch?v=AkBjJ6OunR4&list=WL&index=28

Adding my 2 cents:

I am writing an app that tracks:
Food: Healthy? Tasty? Quantity?
Sleep: Do I Feel Rested Y/N? Quantity? Can cross-ref against sleep trackers
Work day: Productive? Stressful? Hours Overtime? I’m comparing this to RescueTimer logs
Did I Get To A Personal Project Y/N (for example… writing this app :wink: )
Exercise: Morning Stretching Y/N? Yoga Y/N? Cardio? Anaerobic Y/N? Can cross-ref gainst sports tracker
Day Mood? from bad -> great
I can also add a specific event w/ cause - IE My mood NOW is BAD because WORK. This way I can can also identify specific changes in mood. In fact I don’t know if mood-modelling should go from state to state: ie - stay Good until bad, then stay bad till I say otherwise, or if it should somehow reset to neutral at some rate… to study.
Learned something Y/N?
Made time for Loved Ones Y/N?
Grateful Y/N?
Fun Y/N?
Relaxing Y/N?

Rather than go straight to a predictive or even explanatory model, I’m thinking a first step is to understand correlation… especially since I’m not fully sure what I want my dependent variable to be (mood or work/productivity).

Comments/questions welcome!

Sergio, you’re speaking about something I call “emotional momentum”… ie the carryover from past days. There’s a few ways to deal w/ this: 1) you could try modelling w/ two-day averages or 3 day or whatever… 2) you can reintroduce as a new variable yesterday’s stress and use that as a predictor… see what I mean?

1 Like

So I’ve never been able to co-relate happiness / anxiety with anything in particular successfully except with broad factors (such as in a relationship or not) but I do write a qualitative log almost every morning. The idea of knowing what I am feeling is pretty much lost on me.

One thing I will be embarking on next is to co-relate my moods using HRV. HRV has been found to have a very strong correlation with moods and intensity of activity. I think I will be able to make more headway using that for modelling / understanding / predicting happiness and anxiety!

i would like to see the papers on this. I always thought that it was a measure of activity and perhaps general health. Marc

Let me check with the lead author on our modeling paper about sharing findings. It’s under review at the moment and I don’t want to overstep with sharing early.

HRV & mood:
https://www.google.ca/search?q=hrv+and+mood&rlz=1C5CHFA_enCA686CA687&oq=hrv+and+mood

HRV & activity intensity: resting HRV is co-related to your recovery period from your previous activity, and by that, acts as an indicator for how much you can, without injuring yourself, increase the intensity of your current activity.

The complication of HRV comes from the fact that both physical emotional stressors will cause similar changes in HRV. So it is tough to tell why your HRV is high / low but and simpler to decide on a course of action once you you’ve established the state of your HRV.

Sorry I don’t have papers immediately available to me right now but these are well known facts about hrv and exploited by professional athletes on a daily basis. Should not be hard to find the papers!

1 Like

Yes, I agree. With patients with heart disease, chronic disease and age HRV decreases. Measurement is difficult because as you say it is effected by age, weight, exercise, activity, emotional stress, and other factors including how much sleep you had last night. Recovery HRV may have advantages as it may reduce a couple of these factors. As far as the state goes, this was interesting if there was a way to measure it. Maybe like cortisol levels. Baseline HRV, if we could determine such things, would like by a reflection of the health of the autonomic nervous system, but the question is how to do this. Perhaps 24 monitoring of pulse?? Clearly as a short term measurement, there are far too many confounders. Perhaps it can be combined with something else. One suggestion is Blood Pressure through BRS, but this has also similar although fewer issues.

Digging this back up to post an update:

A recent journal paper we have on modeling correlates of mood is now up! https://link.springer.com/article/10.1007/s00779-018-1123-8

You can see the web/phone app (EmotiCal) we built for logging this data + intervention results here: https://people.ucsc.edu/~swhittak/papers/HCI-emotional_examined_life_final_2017.pdf

2 Likes

Hi,

if any of you is interested, I have set up a web app to track daily information about yourself and easily process the correlations between all of it: https://app.getdailys.com/

It’s not going to give you a deep analysis of the causes/consequences that affect your happiness or anxiety, but I believe an easy tool to track correlations is a good start.

It’s free and all so don’t hesitate to give it a shot :slight_smile:
I am planning to improve the app a lot so any kind of feedback is welcome.

I’d like to read more from your end on this subject :slight_smile: - Weekend read

1 Like

@IsaacGerg -

My little experiment in this space posted in this thread:

Honor Band 4 - Experience from using for few days + Flaws & lesser features in iPhone iOS App? Your thoughts on these points - esp if diff on Android?

Having read so much about the (TruSleep & 24/7 HRM) following items I wanted to try my first Fitness Band (instead of Mi Band):

  • That iPhone iOS App is crappy compared to Android and wont even give us ability to do deeper dive into the Charts.
    • I went by the reviews and assumed that iOS App will be same as Android
  • Its irritating to wear the band all day & night
    • After a point my hand/ wrist feels sticky & does not like the feel of the band, especially when resting or sleeping - so it conflicts with wanting to use it for Nap & Sleep monitoring
    • Is there a solution around this?
      • Are Mi Band or Honor Band 3 or other bands any more comfy/ skin/ wearing friendly? Or is this the same for any wrist/ hand based band/ watch tracker?

My experience of Band Features that ARE/ ARE NOT important to me and would like your feedback on my iOS experience (especially if different on Android):

  • Sleep Monitoring - TruSleep
    • It wont do less than 3 hr tracking - So if you wake up or your sleep breaks into 2 naps poof!
    • Unable to Zoom into the data & charts
  • Continuous HRM 24/7 Monitoring & Logging
    • Objective of this was to have 24/7 Log (which it is collecting) and be able to Zoom into how/ when HR changes
      • Time - Activity introspection
    • So maybe at a certain time in the day doing X thing I see a spike (driving in traffic) +
      • Which would help if I could track HR + Location / GPS to see & remember what was causing HR up/ down.
  • Workout / Training HRM
    • I am not big on Running/ Cycling - I do Weight Training & Bodyweight Calisthenics
    • So I turned the “Training Mode” on from the Band (dont care about step counts)
    • I was hoping to see the HR data being logged more intensively for workouts
    • Cant see anything in the iOS App - Totally gone or never logged from Band to Phone or same Android vs iPhone issue
    • MAYBE I should just replace this with a Chest Band and another App to log & view the details?
  • Steps Activity - Dont really care
    • X
    • ​
  • Data Logging / Integration of Huawei Health Data with Apple Health App / Kit (Is it more/ better in Google Health??)
    • Active Energy - ??
    • Heart Rate - Key… But not sure of its granularity and my ability to see it in a Chart
    • Height (kinda static for me)
    • Sleep Analysis - Key… But not sure of its granularity and my ability to see it in a Chart
    • Steps - Not a priority for me
    • Walking + Running Distance - Not a priority for me
    • Weight (using a 2 scales (Eufy + Yunmai) to log but not seeing it charted in Huawei Health App)

Based on the above FLAWS do you think there is a WORKAROUND or BETTER FIT of a Tracker for me?