How to read interaction coefficients in logistic regression?

Asked Dec 27 '22 at 19:17

Active Dec 28 '22 at 05:13

Viewed 30 times

How do i read age:gender1?

Gender1 is male.

Call:
glm(formula = cardio ~ gender + age + gender * age, family = binomial, 
     data = CVtrain)
Coefficients:
               Estimate Std. Error z value Pr(>|z|)

(Intercept)  -1.060e+01  3.328e-01 -31.843  < 2e-16 ***
age           5.164e-02  2.108e-03  24.492  < 2e-16 ***
gender1       4.146e-01  1.758e-01   2.358 0.018379 *

age:gender1  -8.018e-03  3.283e-03  -2.442 0.014597 *

edited Dec 28 '22 at 05:13

kjetil b halvorsen

77,844

asked Dec 27 '22 at 19:17

Antonio

2

How would you read them for a linear model? – Dave Dec 27 '22 at 19:21
Would it be "The effect on the presence of cardio of 1 unit increase in age is exp(-8.018e-03) = 99% more for those that are male as compared to females" ??
By that measure, gender shouldn't be included at all.
– Antonio Dec 27 '22 at 19:53
1

There is no exponential in a linear model. How would you interpret the interaction in a linear model? – Dave Dec 27 '22 at 20:01
The effect on the presence of cardio of 1 unit increase in age is 8.01e-03 less for those that are male as compared to females – Antonio Dec 27 '22 at 21:33
1

So then that’s the effect on the log-odds, right? – Dave Dec 27 '22 at 21:48
1

See here. – dimitriy Dec 27 '22 at 22:36

How to read interaction coefficients in logistic regression?

0 Answers0