Probit, Logit and LPM - Marginal Effect


I am trying to learn about regressions with binary as the dependent variable. I have tried to figure out something but has gotten a little bit lost in my interpretation. Would someone please check if this is correct?

  • With Probit and Logit Models, we only interpret the coefficients as Yes/No, True/Not True. There is not much to interpret in the coefficients other than that.
  • LPM is supposed to give us coefficients that can be interpreted as the percentage probabilty. This often fails, because of the models weakness. Hence, LPM is not a model thats widely used anymore.

I would be extremely grateful for an answer. Thanks.


Active Member
im going 'false' on both bullet points.

The coefficients in logit models are log odds ratios of some form or another. in probit the model coefficients are usually expressed as LD50, as -intercept / beta.