User blog comment:Sirebel/Mathematician Required - Apply Within/@comment-27170954-20170523122715/@comment-27170954-20170523190807

I still believe it ot be the case and hopefully the next data set i colelct proves it one way or another.

This data set had 269 data points, enought to calculate all 26 upgrades independently and i did so. Engine, for example looks like this (E1 to E5)

however, the 95% confidence intervals for all 5 overlap and if your round to two decimals, the range is 0.26-0.29. This would not support a wider range of, say, 0.2-0.4. This is the case for all categories.

The secnd reason i think the levels in acategory will turn out ot be the same is there are only 2 out of 269 errors when i assume they are the same. If i calculate them independenelty, then there are 6 out of 269 errors.

I think if i can get a more balanced data set I might be able to refine the answer. But this is only one car, and i've done so far doesn't prove it either way. Once there is a process we can try it on multiple cars.