I have an idea but I'm not sure if this is 'done' or not. Here's the problem:
I have in hand related surveys of the child protection systems in two Canadian provinces, Ontario and Quebec. In Quebec they have a case diagnosis of "severe behaviour problems" but this case diagnosis does not exist in Ontario. However, there are similar checklists of behaviours in the questionnaires for both provinces. Based on this checklist of behaviors, I can produce, for the Quebec data, a logistic regression model of 4 or 5 behaviours that classifies maybe 87% of cases falling into the "severe behaviour problems" category.
Recall that I have no such category of '"severe behaviour problems" in Ontario but I have all the same markers. I would like to plug all the Ontario cases into the model derived from the Quebec data and extrapolate, based on this model, how many Ontario cases would be categorised as "severe behaviour problems" if, hypothetically speaking, such a category existed in Ontario. Obviously the validity of this procedure would depend on a bunch of assumptions and it's not clear to me how you would give a confidence interval for this extrapolated figure.
David Klein
|