# Analysis of Impact Hours for Retweeting and Attending Meetings

The work that supports this is available [here](https://nbviewer.jupyter.org/github/andrewpenland/praiseanalysis/blob/main/retweets-and-meetings-analysis.ipynb).

## Analysis of Retweeting Impact Hours

I am only looking at data from period 7 (mid-January) to period 17 (mid-May). This is when 1.) the TEC Discord server existed, 2.) the 3-quantifier system was in place.

### Impact Hours for "Retweeting" Praise:
A praise that mentions "retweeting" may be for multiple tweets. All told, the average IH value was about 0.54 Impact Hours per Praise that mentioned re-tweeting. The Impact Hours due to Praise for retweeting accounts for about 4% of all Impact Hours

All "retweeting" praises were intended to receive the same basic value during a Period. Both this basic value, and the total proportion of Impact Hours awarded for retweeting, varied quite a bit by Period.



## Impact Hours for Attending Meetings

Another point of interest was attending meetings (as opposed to hosting or leading). Finding these was harder, since there wasn't a standard keyword like "retweeting".

I made two lists of keywords:

* List1 = ["coming","joining","attending", "attendance", "being", "present", "presence"]
* List2 = ["sync", "call", "meeting", "party", "session", "sesh"]

If a praise contained at least one word from each list, it got counted as "attending a meeting". It's possible that some got missed, this would't catch "Showing up for today's research WG", but it correctly misses "Hosting a party".

The average number of Impact Hours for this set of Praises was 1.14 Impact Hours. This is driven a bit higher by the fact that some of the Praises caught were of the form "Doing X and Y and attending a meeting", where X and Y were objectively more difficult tasks. The median value for attending a meeting was almost exactly 1.0 Impact Hours. I estimate that by attending one meeting per week during this period, a user could expect to earn somewhere between 12 and 20 Impact Hours.

There was again variation of average Impact Hour value by period.


## Estimate

Based on this analysis, I would estimate that about 15% of Impact Hours awarded were due to retweeting or attending a meeting. This is slightly lower than the amount predicted by the hand-labeled data, but not drastically so.

## Final Note: A Word on Comparison and Variability

There is an interesting effect due to the variability of Impact Hour valuations between periods. Consider the largest IH value given for retweeting: during Period 14, praises for retweting received about 0.85 Impact Hours. This ranks higher than 400 meeting-related praises (about 39% of the ones we found). **Some meeting Praise turned out to receive fewer Impact Hours than some retweeting Praise, which was interesting.**