I'm doing a contingency table analysis with one column variable and about
275 row variables. The column variable is the number of specific behavior
problems noted in each kid, from a checklist, zero to 12 - so this is an
ordinal variable. The row variables are all sorts of parameters related to
the child, child's family, and how the child's case was handled by child
protection authorities. These are a mixed bag - mostly nominal and ordinal.
I can get chi-squared values from the contingency table for the relatedness
of each of the case variables to the number of observed behavior problems.
What I want to do, however, is compare this relation across two Canadian
provinces, Ontario and Quebec. For instance, take the relationship between
the number of observed problems and the likelihood that the child is living
with a single mother. What I need is a strategy to compare these
relationships between these provinces, i.e., between different subsets of
the total data base of cases. If possible I would like a method that is
applicable to all the variables uniformly, whether they are continuous,
ordinal, or nominal.
Any suggestions?
David Klein
|