I’ve just generated some graphs on random data (100 iterations for each size) and some of the results look very suspicious:
Precision 10
Precision 7
Precision 13
Since the error rates for smaller and bigger sample sizes look alright, it seems like it might be related to the Bias correction down for medium sized arrays. I’ll try to take a more detailed look in the next few days.
I’ve just generated some graphs on random data (100 iterations for each size) and some of the results look very suspicious:
Precision 10![index](https://cloud.githubusercontent.com/assets/1313584/22527932/579504d0-e8d1-11e6-9e4a-8c421aae5fcc.png)
Precision 7![precision_7](https://cloud.githubusercontent.com/assets/1313584/22527933/5795cbea-e8d1-11e6-9398-5b1cf9a55fdf.png)
Precision 13![precision_8](https://cloud.githubusercontent.com/assets/1313584/22527931/579422f4-e8d1-11e6-9796-7f6421e189cf.png)
Since the error rates for smaller and bigger sample sizes look alright, it seems like it might be related to the Bias correction down for medium sized arrays. I’ll try to take a more detailed look in the next few days.