search

saigegit / SAIGE

Development for SAIGE and SAIGE-GENE(+)
GNU General Public License v3.0
66 stars 29 forks source link

Firth SE #46

Closed bgorm closed 2 years ago

bgorm commented 2 years ago

Hi Wei,

When Firth mode is enabled, the beta estimates appear to be more accurate for some SNPs, but z (beta/se) no longer corresponds to the p-value given in the p.value column. It appears that the Firth z generally corresponds to a smaller p-value than given in the p.value column, suggesting it is not as well-controlled for inflation.

Does it make sense to enable an option to allow the SE to be back-calculated from the SPA test p-values? This would be similar to the option --firth-se in REGENIE to back-calculate the Firth SE from the LRT p-value.

Thanks Bryan

weizhou0 commented 2 years ago

Hi Bryan,

Thanks a lot for the great suggestion! Now in the version 1.1.6, we've added this feature, so SE is back-calculated from the SPA test p-values when BETA is estimated using the Firth approx.

Thanks, Wei

bgorm commented 2 years ago

Thanks Wei!

tom-a-bond commented 1 year ago

Hi, what is the name of the flag to back-calculate the SE from the SPA test p-value and firth BETA please?