search

bvilhjal / ldpred

MIT License
96 stars 57 forks source link

Bug report for computing z scores from p-values #144

Open mxcai opened 4 years ago

mxcai commented 4 years ago

There is a bug on computing z-scores from p-value: In sum_stats_parsers.py, line 182, it computes the lower tail of the normal distribution, which leads to reverse sign of Z-scores. Either adding a negative sign or the replace the ppf(pval_read / 2.0) by (1-ppf(pval_read / 2.0)) can solve the problem

gabraham commented 4 years ago

@mxcai maybe this is related to issue #139 ?

Also, do you know why on the same line the code divides by sqrt(N), it's no longer a z-score in that case (unless this adjustment is taken into account later at some stage).