Question on adjusted P values reported in collapsed output

[mt1022]

Dear developers,

It seems that the adjusted p values in the collapsed output was calculated from filtered p values. Shouldn't multiple testing correction be done before filtering p-values with the default pvalue cutoff of 0.05? To do the multiple testing correction correctly, could I set --pval-cutoff 1 and then filter the output by adjusted_pval < 0.05 manually?

> collapsed[order(-pval_combined)]
      pval_combined adjusted_pval
    1:  4.993873e-02  4.993873e-02
    2:  4.977925e-02  4.978278e-02
    3:  4.968042e-02  4.968748e-02
    4:  4.958320e-02  4.959377e-02
    5:  4.957438e-02  4.958848e-02
   ---                            
14073: 6.443817e-267 1.814192e-263
14074: 3.203795e-276 1.127496e-272
14075:  0.000000e+00  0.000000e+00
14076:  0.000000e+00  0.000000e+00
14077:  0.000000e+00  0.000000e+00

Besides, I noticed that there are two parameters on calculating combined p-values:

  --dependence_test {none,mic,pcc}
                        the method for measuring the dependence between frame1 and frame2. This test could help determine whether the combined p-values should be ajusted to account for the dependence between two test (i.e. F0 vs F1 and F0 vs F2). mic: Maximal Information
                        Coefficient; pcc: Pearson Correlation Coefficient.
  --stouffer_adj {none,nyholt,liji,gao,galwey}
                        the method for adjustment the cominbed p-values to account for the dependence between two tests (i.e. F0 vs F1 and F0 vs F2). see details at: https://search.r-project.org/CRAN/refmans/poolr/html/stouffer.html

What's the recommended way to set the two parameters?

Thanks in advance.

[ericmalekos]

Hi, @mt1022 did you come to a conclusion about this? Are adj pvalues different when you set the cutoff to 1?

[mt1022]

We didn't try this. Combined and adjusted p values are calculated in this script. The code is hard for me to follow.