I am implementing the Westfall-Young (1993) free step-down resampling methodology to correct for multiple hypothesis testing. In my setting, I'm interested in the effects of a regulation that vary over time; thus I'm running several regressions along the lines of

Code:
reg y1 x1 x2 x3 i.qtrs_from_reg_plus8
reg y2 x1 x2 x3 i.qtrs_from_reg_plus8
reg y3 x1 x2 x3 i.qtrs_from_reg_plus8
where qtrs_from_reg_plus8 is defined to be always positive (i.e., this variable would take the value 4, for an observation that is 4 quarters prior to the regulation). I am doing this test using the -wyoung- package by Julian Reif (full citation below, though my question may transcend that particular package). The package (and the Westfall-Young method generally) require specifying a variable of interest for which the multiple-hypothesis-corrected p-values are reported. In my case, that's a factor variable -- that is, not one variable but several. When I run

Code:
wyoung, "reg y1 x1 x2 x3 i.qtrs_from_reg_plus8" "reg y2 x1 x2 x3 i.qtrs_from_reg_plus8" "reg y3 x1 x2 x3 i.qtrs_from_reg_plus8" familyp(qtrs_from_reg_plus8) bootstraps(1000) seed(12345)
The wyoung package fails , with the error message "qtrs_from_reg_plus8 not found".

One alternative is to specify individual values for this variable of interest, i.e.
Code:
tab qtrs_from_reg_plus8, gen(tau)
reg y1 x1 x2 x3 tau1-tau20
reg y2 x1 x2 x3 tau1-tau20
reg y3 x1 x2 x3 tau1-tau20
and then run -wyoung- separately, once for each tau, of course using a pre-set seed so as to use the same set of bootstrap draws for the MHT-corrected value for each tau. Is there a better way? (Also, any comments on whether it is econometrically sound?)

Many thanks.
Robert

Citation for -wyoung- package: Jones, D., D. Molitor, and J. Reif. 2018. "What Do Workplace Wellness Programs Do? Evidence from the Illinois Workplace Wellness Study." National Bureau of Economic Research Working Paper No. 24229.