I am using logistic to see the relationship between churn and minute seen.
I get something like this
> summary(log_model_test ) Call: glm(formula = isChurn ~ avg_minutes_seen, family = binomial(link = "logit"), data = train_churn_2) Deviance Residuals: Min 1Q Median 3Q Max -0.7142 -0.5852 -0.4712 -0.2987 4.8138 Coefficients: Estimate Std. Error z value Pr(>|z|) (Intercept) -1.2361998 0.0089931 -137.46 <2e-16 *** avg_minutes_seen -0.0005472 0.0000062 -88.26 <2e-16 *** --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 (Dispersion parameter for binomial family taken to be 1) Null deviance: 209628 on 285696 degrees of freedom Residual deviance: 198453 on 285695 degrees of freedom AIC: 198457 Number of Fisher Scoring iterations: 6
Now, I am trying to interpret results; however, I don't know what is the correct way to express this relationship. Which is the right way to interpret the result? I just want to say how much impact minutes seen has on churn.
This fitted model says that, the odds of churning (churn = 1) over the odds of not churning (churn = 0) is exp(-0.0005472) = 0.9994398
1- For a one-unit increase in the average minute seen, the expected change in log odds is 0.0005604
2- For every one unit change in an average minute seen, the log odds of non-churn (versus churn) increases by 0.999766
3- So we can say for a one-unit increase in an average minute seen, we expect to see about (1 - 0.9997661) x 100 ) = 0.05602% decrease in the odds of being in churn class.
Thanks in advance