FDR analogous confidence intervals

[Sebastian Baumeister]

Dear Statalist:

Is there any possibility to derive adjusted confidence intervals that are comparable to FDR (Benjamini Hochberg false-discovery-rate)-adjusted p-values in Stata? I used Newson's qvalue.ado to calculate qvalues but I cannot figure out how to get equivalent CIs. I found R (https://bmcbioinformatics.biomedcent...71-2105-12-288) code but I would like to in Stata.

Specially,
I run multiple Cox models for different exposures and effect-modifiers and get HRs for combinations of exposures and modifiers. In addition to providing qvalues for the interaction term i would like to get FDR-compareable CIs from lincom.

foreach exp of varlist x1 x2 {
foreach mod of varlist m1 m2 {
quiet stset age_dx_cens, f(ovaryca==1) origin(time ENTRY_AGE)
quiet stcox c.`exp'##i.`mod' $m3, strata(age_5) nolog
parmest, saving("$ovarian_int\`exp'_`mod'", replace)
di "***************HR by levels of modifier***"
levelsof `mod', local(levels)
foreach l of local levels {
lincom c.`exp' + c.`exp'#`l'.`mod', eform level(95)
}
}

Any suggestions? Thanks in advance Sebastian

[Sebastian Baumeister]

Update: Stata does not allow FRD adjusted CIs but Sidak-adjusted CIs. The latter can be computed using the parmcip module of the SSC package parmest.

[Federico Tedeschi]

Given false coverage-statement rate-adjusted Benjamini Hochberg confidence intervals are CIs with a coverage (for each single parameter) different from 0.05, what I do is just to use the "level" option to change the CI coverage.