,
10 tweets,
5 min read
Read on Twitter
Proper analyses of Whitehouse et. al.’s data reveals that moralizing god PRECEDE the rise of complex societies, but serious data concerns persist. psyarxiv.com/jwa2n @babeheim @DrQueue @martinlangcz @willismonroe @mmuthukrishna @slingerland20 @RachelASpicer @aiyanakoka
To see the issue, first take a look at the data the authors’ analyzed. 1 = moralizing god present, 0 = evidence of no moralizing gods, and NA = Missing data. Data for Whitehouse et. al.’s 12 key location (green when writing appears) 
All the data leading up to the appearance of the moralizing gods are MISSING, except in the Yellow River (which is an error in Seshat and should be a ‘1’). How did Whitehouse handle this missing data?
Whitehouse et. al. converted all of their ‘NA’s (missing data) to ‘0’. The recoding, which turns an absence of evidence into evidence for absence, transformed 61% of the values of the dependent variable cases and 98% of the ‘0’s (no moralizing gods) are converted NA!
Using the same analytical approach as the authors, we re-analyzed the data by retreating the NAs as missing data (left). We then used the results to predict the presence or absence of moralizing gods for missing data (NAs), which occur mostly at low levels of social complexity.
This show moralizing gods PRECEDED the dramatic increase in social complexity (prior figure, right). @sciam
In some analyses, Whitehouse et. al. only used the date of 1st appearance for moralizing gods & looked at the growth of social complexity before & after. But, they didn't consider forward bias. What if under those 'NA's are actually MGs--1st appearance dates are biased forward?
To explore this, we moved their first appearance dates back by 1 century, the smallest time unit in their data. This REVERSES the results: social complexity grows faster after compared to before. @NYTScience
Are the dates of first appearance for moralizing god forward biased? The above analyses say they are forward biased by at least 400 years. One key factor is writing: Whitehouse et. al.'s data show that, usually, moralizing gods appear just when writing arrives. @minzlicht
Finally, for a detailed review of the vetting and data quality in Seshat, see psyarxiv.com/2amjz
