Dear Statlist community,
I have a question regarding the complete precedere of a probit analysis.
I conducted a probit regression analysis if a HH sends a migrant based on HH characteristics such as HH_Size, Agricultural Income, Education and so on. I cleaned my data, run the probit regression and have my outputs.
I am wondering if (or which) tests or check ups I need to conduct to achieve a robust output. I read through some post online and I found some options I can do but they dont seem compulsory to me.
What i found is:
Interaction effects: check for interaction effects between the variables. If the interaction effect is significant I should incorporate it in my model. But this results just in a another model spezification I guess.
Linktest: Checks if model is correctly spezified. If i run the linktest and the result shows its not correct specified, I should use these model outputs?
Distribution: my metric variables are not normaly but chi2 distributed. Is this a problem?
Edogeneity: I would guess Agr_Income is influenced by HH_Size. Do i control for this by the interaction effect? Or how can i check for this relation?
Homoscedasticity: Do i need to check for this?
Are there other tests, characteristics I need to check?
Thank you in advance!
0 Response to Probit regression procedure
Post a Comment