Clustered t-test ----------------
^clttest^ outvar [^if^ exp] [^in^ range] , ^cluster(^clustvar^)^ ^by(^byvar^)^ ^strata(^stratavar^)^
Description ----------- This program calculates a cluster-adjusted t-value for comparing continous outcomes when the unit of randomization is a cluster. It calculates the adjusted t-value based on the effects of clustering by ^byvar^ group. See reference below for formulae. The results include cluster adjusted means and confidence intervals for each group; the difference between these; and, for stratified data, the variance weighted mean difference. The ICC is calculated separately for each group.
^outvar^ is a continous outcome variable. It must be clustered within ^clustvar^. ^clustvar^ is a categorical variable denoting clusters of observations. It is ^not^ optional. ^byvar^ is a dichomotous variable denoting groups to be compared. It is ^not^ optional.
^stratavar^ is a categorical variable denoting strata of groups.
Note: This program does not yet support comparison of two variables analogous to the ttest var1=var2 command.
. ^clttest score, cluster(school) by(control)^
where observations of ^score^ are clustered within ^school^, and the comparison is made between the two groups identified by ^control^.
Also see -------- ^clchi2^
Jeph Herrin Yale University email: jeph.herrin@@yale.edu
Reference --------- Donner A, Klar N. Design and Analysis of Cluster Randomization Trials in Health Research. Arnold, London. 2000