In the context of my PhD, I'm doing a Fit model using stepwise method to determine the set of factors that best explain my variable of interest (energy balance). When I click on GO this alert message appear: ALERT: Failed to converge (step-halvinglimit). I have put 50 variables in the construct model effect box. Then I clicked on RUN. I choose my stopping rule and direction and clicked on GO. Did I miss something? See photos attached. Thanks for your help.
I see that 5848 rows are excluded. What is the basis for excluding those rows? How many rows are included?
I see that your response, energy_BEE, is ordinal. How many levels? How are the levels determined?
This response leads to the step-wise procedure for ordinal logistic regression with cumulative logits by default. Is that arrangement reasonable for your data and modeling?
Are the predictor variables collinear for the included rows?
Have you explored the distribution of the response and predictor variables for the included rows to see if you have a separate issue or any other problem with the data used for the ordinal logistic regression?