Jon Heron posted on Friday, August 18, 2006 - 8:15 am
I've derived a 5-class LCA based on 5 binary observed variables
I would like to use this 5-level factor as a predictor in a logistic regression model.
If I include this binary outcome as a 6th variable in my usevariable list then I get a set of odds ratios as if the class-variable were the outcome and my binary variable the predictor, which is equivalent to the unadjusted estimate I seek thanks to odds ratios being reversible.
However, I want to incorporate other covariates into this model as confounders so then I can no longer employ the method above.
How can I model a multivariable logistic model in this way?