Welcome! We're just getting started, so check out the blog often and suggest and vote on features!

Blog

How Metrics Affect Each Other

by Kevin on Friday, April 18, 2008

I’m excited to announce a brand new feature on lifemetric: relationships. Relationships are the calculated similarities from one metric to another. They are found on the right sidebar of your home page and user pages:

Let me take a moment to explain these results:

  • My mood is directly proportional to the amount of sleep I get
  • My mood is inversely proportional to the stress in my life
  • My stress is inversely proportional to the amount of sleep I get

Generally the results can be interpreted as such:

Coefficient Interpretation
-1.0 to -0.7 strong negative association
-0.7 to -0.3 weak negative association
-0.3 to +0.3 little or no association (filtered from view)
+0.3 to +0.7 weak positive association
+0.7 to +1.0 strong positive association

We’re working on making these results a little bit easier to interpret at a glance, but the inital results have been quite intriguing.

(For the statistically-inclined, these are Pearson product-moment correlation coefficients, calculated from each pairs of metrics you are subscribed to and update on a regular basis.)

Comments

LostxVampire on Thursday, October 09, 2008 at 4:03 PM

It seems like the problem is fixed, because they’re shwoing up again! :)

LostxVampire on Monday, September 29, 2008 at 4:24 PM

I can’t see any of the relationships between them. On my metrics page there’s a graph and the word ‘relationships’ next to it, but except for the title there’s just empty space… So does anyone know how I can fix this?

mblarsen on Monday, April 28, 2008 at 1:04 AM

@Kevin: Hehe, thanks for the explaination – I’ll let you work out the math. Keep up the good work :o).

Btw: I know it is a different area, but maybe this can give you some inspiration as well http://en.wikipedia.org/wiki/Slope_One#Item-based_collaborative_filtering_of_purchase_statistics

Kevin (OP) on Friday, April 25, 2008 at 1:58 PM

@mblarsen: You’re just really predictable! JK… The algorithm looks at your last 180 days worth of entries as a whole, and discards days in which you didn’t have an entry for all metrics. Since you have so many, you probably don’t update each and every one every day (which is fine, just not something I thought of…) I’ll have to tweak that a bit.

mblarsen on Friday, April 25, 2008 at 4:12 AM

Hmm.. All my values are either perfect positive or negative. Doesn’t this seem highly unlikely?