st: Repeated measures testing

["Nick Winter" <nwinter@policystudies.com>]

Greetings,

In some survey data, I have three variables that measure whether a given
school improvement strategy is used with each of three types of schools,
A, B, and C.  (Variables d1a, d1b, d1c -- d1a codes as a 0/1 whether or
not strategy one is used with school type A; d1b records the same thing
for school type b, and so on).

I'm interested in testing whether a strategy is used differentially for
different types of schools.  If there were two school types, this would
be easy -- simply -ttest- that the two variables are equal:

	. ttest d1a = d1b

But with three, it gets more complex.  

One approach, since this is survey data, is to first use -svymean- to
get the means of all three, then test jointly that a=b and a=c:

	. svymean d1a d1b d1c, complete
	. svytest d1a=d1b, notest
	. svytest d1a=d1c, accum

Another approach would be to set it up as a repeated measures ANOVA --
reshaping to long format, and treating the three types as a repeated
measure.  But with thousands of cases, this seems cumbersome at best.

Thoughts?

Thanks,
Nick Winter


-----------------------------------------------------------
 Nicholas Winter, Ph.D.                     P 202.939.5343
 Policy Studies Associates                  F 202.939.5732
 1718 Connecticut Avenue, NW     nwinter@policystudies.com
 Washington, DC 20009-1148           www.policystudies.com
-----------------------------------------------------------

<<winmail.dat>>