1/ Analyses of precinct data in NH suggest risk that fractional counting by machine may have moved results by 3-5% of gap between Sanders and Buttigieg. @jvgraz recently discussed 2016 suspicions per Stanford study (see also my work in CP + w/EJUSA and @LuluFriesdat led study). 
2/ We cannot draw a conclusion of suspicion about movement from precincts of 349 or less to 350-1999 for Sanders v. Buttigieg. Comparing hand v. scanner counts? Sanders just did better in smallest townships. (Klobuchar diff. *is* suspicious per multivariate regression analysis.)
3/ I worked with an Ivy League trained prof. who uses quantitative and qualitative data in his research. We controlled for several factors like wealth.
Our data frame: easyupload.io/kwg3v4
Wealth, importantly, only explains about 10% of the variance by precinct size:
Our data frame: easyupload.io/kwg3v4
Wealth, importantly, only explains about 10% of the variance by precinct size:
4/ There are important factors beyond wealth and precinct size, however. Median age, Sanders' 2016 share, and 2019 donors by campaign were looked at. These factors help with additional explanation for the variances, but not completely. See especially those highlighted in red.
5/5 Data based tests only raise suspicions and point to further investigation. THIS IS NOT PROOF OF MALFEASANCE. We would need to review paper ballots.
But NH refuses calls for outside Risk Limiting Audits & bars public/media from scrutiny (other states allow).
Yes! I tried.
But NH refuses calls for outside Risk Limiting Audits & bars public/media from scrutiny (other states allow).
Yes! I tried.
