I have an ordinal outcome variable with the following levels (unlikely, somewhat likely, very likely) that I want to test in an adjusted regression model with about 20 predictor variables (including continuous, categorical and binary type). I ran a cumulative logistic regression but the proportional odds assumption is violated. I wanted to know if I can still use Cumulative Logit Regression Technique? I read from multiple sources "When using a proportional odds model, the test of the proportional odds assumption frequently rejects the null hypothesis that proportionality can be assumed. The test can best be described as anti-conservative and nearly always results in rejection of the proportional odds assumption, particularly when the number of explanatory variables is large, the sample size is large, or if there are continuous predictors in the model (Allison, 1999; Brant, 1990; O’Connell 2006)." Does this seem like a doable thing?
Additional info- 1. Partial Proportional Odds model seems like an obvious alternative, however, the problem is that I will have to test each of the predictor separately for the assumption. And this is not the only model I'm running. I have to run 10 other models of similar type and with the same limitation. 2. I cannot combine the levels of the outcome because it will change the meaning of the responses.
Please help! Thanks.
