Logistic regression


Fortran must die
Only the last variable is statistically significant (_cons). Its basically impossible to interpret substantively slopes in logistic regression other than their sign and if they are significant. They refer to a change in the logit for a one unit change in X controlling for other predictors. Not many understand what that means :)

You need to request odds ratios.

Everyone is stupid here except Dason... and he is a robot :(


Active Member
Hi Noesi, Arndim

The slopes and the intercept are not so intuitive like linear regression but can still be interpreted.
The screenshot is not so clear. under the assumption that mer_ng_ml is the intercept. (or just for the example)


When all the values of the predictors (Xj) are zero:
The odds of Y(1) in comparison to Y(0) is: 1.09825955

One unit increase in X1:
Will increase the odds of Y(1) in comparison to Y(0) by 9.8% (a.k.a. the odds will be multiplied by 1.09825955).