My model for thesis is based on CRedit Rating of the firms and its categorized from AAA to CC from 8-1 (scale). The base paper states to use ordered probit regression and other reference paper says ordered logit model. I don't know any of the models. Can anyone guide me when to use such models and do we need to have all variables categorical ?
Regression
- Thread starter abbyzah
- Start date
- Tags regression analysis
Hi,
First, when having ordinal DV with 8 values you probably will get reasonable results also with the linear regression. (the recommended is still the
ordinal logistic regression) (https://www.statisticalanalysisconsulting.com/how-to-analyze-likert-type-dependent-variables/)
I assume the ordered logit model is - the ordinal logistic regression uses the logit function (the logit is the Ln of the odds, the odds is the ratio: p/(1-p).
Anyway, I think that in both models you mentioned the DV is ordinal and both models get similar results. So both are Okay.
The IVs are preferable to be continuous like in a linear regression or using a dummy for categorical.
When the DV is categorical but not ordinal you should use multinomial logistic regression.
First, when having ordinal DV with 8 values you probably will get reasonable results also with the linear regression. (the recommended is still the
ordinal logistic regression) (https://www.statisticalanalysisconsulting.com/how-to-analyze-likert-type-dependent-variables/)
I assume the ordered logit model is - the ordinal logistic regression uses the logit function (the logit is the Ln of the odds, the odds is the ratio: p/(1-p).
Anyway, I think that in both models you mentioned the DV is ordinal and both models get similar results. So both are Okay.
The IVs are preferable to be continuous like in a linear regression or using a dummy for categorical.
When the DV is categorical but not ordinal you should use multinomial logistic regression.
Last edited:
