In precincts where Diebold machines are used, Hillary gets a 7 point advantage over Obama.
According to a best-fit linear model of the effect of precinct size and voting method, voting method accounts for about 4.6 points with high statistical confidence.
That's 4.6 points not explained by precinct size, which is included in the model.
Including variables such as number of Republican voters or other candidates does not significantly alter this result.
Hence the debunking that "Diebold use correlated with large cities correlated with Hillary supporters, thus no conspiracy" is insufficient and further analysis is warranted.
Especially given the wild poll discrepancies.