kingboomie / open-source-human-analysis

simple parsing, hack cleaning, dumb quick analysis. all for https://github.com/open-source-person/gamithra_mood

open-source-human-analysis

simple parsing, hack cleaning, dumb quick analysis.
all for gamithra mood

TOC

• results
  • Linear regressions
  • Pairwise (pearson) correlations
  • PCA

results

Cumulative change

Plot goes up when the measure is on the positive side (> 5), down when on the negative side (< 5). measure == 5 is straight.

Linear regressions

Pairwise (pearson) correlations

PCA

PCA's first component explains about 39% of the variance, second about 18%, and it goes down from there.
This does mean that just using PCA isn't a very good indicator of what's going on here, but this is supposed to be quick and dirty so ¯\_(ツ)_/¯

Inside that 39%, we can check how much different features help explain all other features:

feature	importance
health	0.412
wellbeing	0.39
generosity	0.348
present	0.33
belonging	0.32
gratitude	0.308
gratification	0.26
focus	0.239
independence	0.209
self-worth	0.203
future	0.131
past	0.08

There are two things we can gleam from that table above:

• if we had to only choose one stat to track health would make the most sense. that also means that improving health would give the biggest returns on improving all stats.
• the feelings about the past or the future don't seem to matter too much in explaining all other stats

For reference, here's the table for the second component:

feature	importance
belonging	0.49
gratification	0.414
gratitude	0.39
health	0.36
generosity	0.304
independence	0.28
focus	0.206
present	0.201
self-worth	0.14
past	0.091
future	0.056
wellbeing	0.012