Interpreting Regression Coefficient in Logistic Regression

[filipposebastio]

Hi there,

One question about the interpretation of regressors and their coefficients in multiple regression models. Supposing that I have the following logistic regression

Y = a + b1height + b2weight + e
where y is either 1 = male or 0 = female h is height g is weight

and I estimate my parameters to be:
Y = 0.1 + 0.25height + 0.45weight + e

Now I want to estimate the probability of my observation 'i' with height 175 cm and weight 82 kilos - how would I exactly go about it?
I understand I cannot just plug in my number because the coefficients are interpreted as the change of Y based on the change of 1 unit of X keeping all the other controls constant. Is there another way to go about it?


Thanks!

 

[hlsmith]

I know this isn't the answer you want, but in some software you can use estimate statements to get them or you can add an observation row into you data, with the outcome value missing - so it will be listwise deleted when creating the model, but if you request a predicted probability it will generate one. I post these software based options since they can also generate your 95% CI on the prediction.

P.S., Just in case the above is your model, height and weight are likely highly correlated and also don't predict gender, but gender predict them

 

[filipposebastio]

I know this isn't the answer you want, but in some software you can use estimate statements to get them or you can add an observation row into you data, with the outcome value missing - so it will be listwise deleted when creating the model, but if you request a predicted probability it will generate one. I post these software based options since they can also generate your 95% CI on the prediction.

P.S., Just in case the above is your model, height and weight are likely highly correlated and also don't predict gender, but gender predict them

Hi Hlsmith,

Thanks for your help.

Can I do that in Stata and how?

Thanks!