4 ms·
You don't have a lot of data- but the data is clearly bimodal (Hartigan dip test value: 0.1875, which is "large"- so this is a very unlikely pattern from a unim
by jmount 16y ago
You don't have a lot of data- but the data is clearly bimodal (Hartigan dip test value: 0.1875, which is "large"- so this is a very unlikely pattern from a unimodal distribution). The question that can't be answered from this data alone- is the different sub-population of 1-stars due to something in your app (like a bug only they encountered) or something about them (older iPhone, or competitors entering harmful scores).
- jmount 16y agoOkay that was the "stat" answer (which brings in too many irrelevant details). The "business" answer is think in terms of something like http://en.wikipedia.org/wiki/Net_Promoter http://en.wikipedia.org/wiki/Net_Promoter and say "the middle scores are just noise- so the only meaningful summary you can pull out is the net difference in 4 and 5 star reviews against 1 and 2 star reviews."