Ricardo Ovaldia asked about a pre-post randomized parallel-group trial
design with a binary outcome variable and Kieran McCaul suggested a
repeated-measures (conditional logistic regression) approach.
I have three follow-up questions:
1. Which is better for binary outcomes, Kieran's repeated-measures approach
or an ANCOVA-like approach using the pretreatment values as a baseline
covariate in conventional logistic regression? The do-file below suggests
that completely different conclusions would be drawn from the same dataset
depending upon which approach is used to analyze it.
2. As Kieran mentioned, the repeated-measures approach drops one of the
"main effects" (treatment) so that the model ends up having an interaction
term in it when one of the component "main effects" terms contributing to
the interaction is not in the model. This would be a no-no from what I've
heard, at least for the analogous situation in ANOVA. But, I assume that
this *not* a problem for conditional logistic regression due to the
conditioning. Is that correct?
2. When using the likelihood-ratio test (-lrtest-), which is the proper
model against which to compare for testing individual "main effects" of
treatment and baseline--the saturated model (*with* the interaction) or the
partially reduced model (*no* interaction term, i.e., the model that
includes only both of the main effects)? Or should we be testing a
constant-only model against one with the "main effect" in order to test that
"main effect"?
Joseph Coveney
----------------------------------------------------------------------------
clear
set more off
set seed 20040129
* Fictional dataset based upon Ricardo's description
set obs 328
generate byte intervention = _n > _N / 2
generate byte baseline = 0
display 328 / 2 * 0.29
replace baseline = 1 in 1/66
display 328 / 2 * 0.40
replace baseline = 1 in 165/212
tabulate intervention baseline, row
generate float ord = uniform() // no particular correlation
sort intervention ord
drop ord
generate byte followup = 0
replace followup = 1 in 1/85
replace followup = 1 in 165/259
tabulate intervention followup, row
* ANCOVA-like approach
xi: logistic followup i.baseline*i.intervention
estimates store A // Saturated model
xi: logistic followup i.baseline i.intervention
estimates store B // "main effects" only model
lrtest A B, stats
xi: logistic followup i.baseline
lrtest A ., stats // This one?
lrtest B ., stats // Or this one?
xi: logistic followup i.intervention
lrtest A ., stats // This one?
lrtest B ., stats // Or this one?
* Repeated-measures approach
generate int personid = _n
rename baseline seatbelt0
rename followup seatbelt1
reshape long seatbelt, i(personid) j(period)
xi: clogit seatbelt i.intervention*i.period, group(personid)
estimates store C
xi: clogit seatbelt i.period, group(personid)
lrtest C ., stats
exit
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
