Joseph Wagner wrote:
Thank you Nick. That was the problem. I now get:
Nonparametric test for trend in hlths
with eval, stratified by id
z = 0.30, chi-squared(1) = 0.09, P>|z| = 0.7617
which confirms my other results.
--------------------------------------------------------------------------------
Results from the do-file (taken, I thought, from the table that Joseph posted)
are summarized below. I've reposted the do-file modified to include the t
test. I can privately send Joseph the log if he has trouble running it. (It's
not clear why -gllamm- doesn't see any observations in his attempts.)
The parametric methods (-gllamm- and -mvtest-/-ttest-) are slightly more
powerful than the nonparametric methods, but all would lead to the same
conclusion.
-ologit, -cluster()- tests a different hypothesis, so I wouldn't expect that it
would give the same results. I haven't examined the proportional odds
assumption. It might also contribute to the difference, but if so I would have
expected that to show up also in -gllamm-'s results to some extent. Roger
Newson's -somersd, cluster()- agrees with -ologit, cluster()-.
Stata command P
-------------------- ----
-signtest- 0.58
-signrank- 0.51
-npt_s- 0.52
-gllamm-, Wald 0.48
-gllamm-, LR 0.48
-mvtest- or -ttest- 0.49
-somersd , cluster()- 0.73
-ologit, cluster()- 0.73
In any event, it seems that the answer to Joseph's original post's question is,
"Yes. Philip Ender's -mvrepeat- is appropriate for use here."
Joseph Coveney
clear
set more off
input byte sco0 byte cou1 byte cou2 byte cou3 byte cou4 byte cou5
1 28 18 6 0 0
2 21 78 44 1 0
3 7 34 96 5 1
4 0 3 18 16 0
end
reshape long cou, i(sco0) j(sco1)
log using A:\HealthScores.log, text replace
tabulate sco0 sco1 [fweight = cou], gamma taub
drop if cou == 0
expand cou
drop cou
mvrepeat sco0 sco1 // or . . .
ttest sco0 = sco1 // or . . .
generate byte del = sco1 - sco0
ttest del = 0 // I believe how Prof. D'Agostino couched it.
drop del
signtest sco0 = sco1
signrank sco0 = sco1
generate int pid = _n
reshape long sco, i(pid) j(tim)
somersd tim sco, cluster(pid)
ologit sco tim, cluster(pid)
npt_s sco, by(tim) strata(pid) nodetail
version 7: ordplot sco, by(tim)
gllamm sco tim, i(pid) family(binomial) link(ologit)
estimates store A
gllamm sco, i(pid) family(binomial) link(ologit)
estimates store B
lrtest A B, stats
log close
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/
