Power calculation- cross-over design

[Natalie Malek]

Hello,

Does anyone have experience with power calculations in STATA for a crossover design? Trying to calculate minimal incidence rate ratio and not sure if there is an existing command to do so? Any advice is appreciated.

Thank you!

[Joseph Coveney]

Not sure about an existing command, but if not, then you can do something like the following. It's for a one-way crossover, but you can embellish the dataset (and model) with sequence assignment (AB/BA), assumptions as to carryover, and whatnot.

.ÿ
.ÿversionÿ16.0

.ÿ
.ÿclearÿ*

.ÿ
.ÿsetÿseedÿ`=strreverse("1527824")'

.ÿ
.ÿprogramÿdefineÿsimem
ÿÿ1.ÿÿÿÿÿÿÿÿÿversionÿ16.0
ÿÿ2.ÿÿÿÿÿÿÿÿÿsyntaxÿ,ÿn(integer)ÿ[Delta(realÿ0)]
ÿÿ3.ÿ
.ÿÿÿÿÿÿÿÿÿdropÿ_all
ÿÿ4.ÿÿÿÿÿÿÿÿÿsetÿobsÿ`n'
ÿÿ5.ÿÿÿÿÿÿÿÿÿ
.ÿÿÿÿÿÿÿÿÿgenerateÿlongÿpidÿ=ÿ_n
ÿÿ6.ÿÿÿÿÿÿÿÿÿgenerateÿdoubleÿpid_uÿ=ÿrnormal()
ÿÿ7.ÿ
.ÿÿÿÿÿÿÿÿÿexpandÿ2
ÿÿ8.ÿÿÿÿÿÿÿÿÿbysortÿpid:ÿgenerateÿbyteÿtimÿ=ÿ_nÿ-ÿ1
ÿÿ9.ÿ
.ÿÿÿÿÿÿÿÿÿgenerateÿdoubleÿxbÿ=ÿpid_uÿ+ÿtimÿ*ÿ`delta'
ÿ10.ÿÿÿÿÿÿÿÿÿgenerateÿlongÿoutÿ=ÿrpoisson(exp(xb))
ÿ11.ÿ
.ÿÿÿÿÿÿÿÿÿxtpoissonÿoutÿi.tim,ÿi(pid)ÿreÿnormal
ÿ12.ÿÿÿÿÿÿÿÿÿtestÿ1.tim
ÿ13.ÿend

.ÿ
.ÿprogramÿdefineÿscanem
ÿÿ1.ÿÿÿÿÿÿÿÿÿversionÿ16.0
ÿÿ2.ÿÿÿÿÿÿÿÿÿsyntaxÿ,ÿn(integer)ÿ[Reps(integerÿ300)ÿDelta(realÿ0)]
ÿÿ3.ÿ
.ÿÿÿÿÿÿÿÿÿlocalÿwhichÿ=ÿcond(`delta',ÿ"Power",ÿ"Testÿsize")
ÿÿ4.ÿÿÿÿÿÿÿÿÿquietlyÿsimulateÿpÿ=ÿr(p),ÿreps(`reps'):ÿsimemÿ,ÿn(`n')ÿd(`delta')
ÿÿ5.ÿÿÿÿÿÿÿÿÿgenerateÿbyteÿposÿ=ÿpÿ<ÿ0.05ÿifÿ!mi(p)
ÿÿ6.ÿÿÿÿÿÿÿÿÿsummarizeÿpos,ÿmeanonly
ÿÿ7.ÿÿÿÿÿÿÿÿÿdisplayÿinÿsmclÿasÿtextÿ"Nÿ=ÿ"ÿ`n',ÿ"`which'ÿ=ÿ"ÿ%04.2fÿr(mean)
ÿÿ8.ÿend

.ÿ
.ÿ
.ÿ//ÿIRRÿofÿ1.25
.ÿforvaluesÿnÿ=ÿ150(50)250ÿ{
ÿÿ2.ÿÿÿÿÿÿÿÿÿscanemÿ,ÿn(`n')ÿd(`=ln(1.25)')
ÿÿ3.ÿ}
Nÿ=ÿ150ÿPowerÿ=ÿ0.74
Nÿ=ÿ200ÿPowerÿ=ÿ0.88
Nÿ=ÿ250ÿPowerÿ=ÿ0.92

.ÿ
.ÿ//ÿTestÿsize
.ÿscanemÿ,ÿn(200)
Nÿ=ÿ200ÿTestÿsizeÿ=ÿ0.04

.ÿ
.ÿexit

endÿofÿdo-file


.

After honing in, you can up the repetitions to get more precise estimates with the chosen delta and sample size.