topic Re: Marginal effect of logistic regression in New SAS User

Marginal effect of logistic regression

VincentTang — Tue, 24 Sep 2019 14:14:20 GMT

Dear all,

Nice to meet you all.

Recently, I met a problem in the estimation of logistic regression.

After running the logistic regression, the estimated coefficients are obtained. However, they are the coefficients of the log odds ratio. May I know is there any codes for finding out the marginal effects.

I have also attached my code here for your reference.

proc logistic data=edu2016;
class Ethnicity Age Gender / param=ref ref=first;
model PrimaryCompleted/Total = Ethnicity Age Gender Ethnicity*Age Ethnicity*Gender Age*Gender/ lackfit;
run;

Thank you very much for your help.

Re: Marginal effect of logistic regression

PaigeMiller — Tue, 24 Sep 2019 14:24:16 GMT

@VincentTang wrote:

Recently, I met a problem in the estimation of logistic regression.

After running the logistic regression, the estimated coefficients are obtained. However, they are the coefficients of the log odds ratio. May I know is there any codes for finding out the marginal effects.

By marginal effects, do you mean the slope of the probability of Y as X changes (instead of the slope of log-odds ratio of Y as X changes)??

Re: Marginal effect of logistic regression

VincentTang — Tue, 24 Sep 2019 14:25:45 GMT

Hello, Paige

Yes. the slope of the probability of Y as X changes.

Thank you for your help

Re: Marginal effect of logistic regression

PaigeMiller — Tue, 24 Sep 2019 14:30:23 GMT

The slope of the probability of Y is not a linear function of X.

The formula looks like this (from Wikipedia at https://en.wikipedia.org/wiki/Logistic_regression) for a regression with two x-variables and an intercept

$where the lower case b is the base of the logarithm used, which is usually e.$