Clustered Chi-squared tests ---------------------------
^clchi2^ eventvar groupvar [^if^ exp] [^in^ range] , ^cluster(^clustvar^) str > ata(^stratavar^)^
Description ----------- This program calculates cluster-weighted chi-square values for comparing dichomotous event rates when the unit of randomization is a cluster. It calculates two adjusted chi-squared values, one based on the effect of clustering pooled across all study groups, and one based on the effects of clustering by study group. See reference below for formulae for the 2x2 case; this program makes the natural generalization to the 2xk case.
^eventvar^ is a dichomotous event variable. It must be clustered within ^clustvar^. ^groupvar^ is a categorical variable indicating comparison groups. ^clustvar^ is a categorical variable denoting clusters of observations. It is ^not^ optional. ^stratavar^ is a categorical variable denoting strata of clusters.
In addition to reporting the clustered chi-squares and their P-values, ^clchi2^ reports Pearson's chi-squared value (with P-value) and the inter-^clustvar^ correlation. If ^stratavar^ is specified, a Mantzel- Haenzel statistic is calculated.
Note: The adjustment for clustering uses (and reports) the intra-cluster correlation estimated by ^loneway^, which applies a correction for imbalanced clusters.
Example -------
. ^clchi2 passfail group, cluster(school)^
where observations of ^passfail^ are clustered within ^school^.
Also see -------- ^clttest^
Author ------
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