Logit Model Interpretation of results

[Ajay Joshi]

So I am running a logit model with categorical independent variables, but I was told to use the i. prefix to split the categories up by my supervisor (this is for an empirical research project). I am measuring the effect an individual's perception on equality in educational opportunities has on the impact of socioeconomic background on likelihood of degree attainment.

However, each categorical variable has 10 categories and I have about 6 of those variables, so I am not sure on whether to include all 60 different categories in my results interpretation for my write up, or which categories I should include as all 60 seems excessive and will take up to many words to explain and include in a table of results. It makes sense to only include the significant ones, but even at that there are still many categories.

I am also running three specifications of this model which means that there could be a lot of categories to include in the table, but have only included screenshots of the final one.

I am also not sure as to whether I need to use the marginal effects to interpret my results or not Aswell?

Any help would be greatly appreciated!

[Richard Williams]

Are the categorical variables ordinal variables? If so consider treating them as continuous. See

https://www3.nd.edu/~rwilliam/xsoc73...ndependent.pdf

[Ajay Joshi]

Hi Richard,

No, they are all categorical variables with distinct categories.

Do you know how I could go about interpreting my results?

Would I be best using odds ratios, normal logit coefficients or marginal effects to interpret my results?

[Richard Williams]

I might be tempted to combine some categories if that seemed reasonable. I would use marginal effects myself.