00:00 Intro 01:11 Binary Response 03:55 Why we can't use linear regression for binary outcomes 06:21 Make a regression model work for a binary outcome 07:50 Visual explanation of logit link function 10:28 Characteristics of logits and probabilities 12:07 Example#1: what is the prob. of a normal birth weight 14:04 Interpretation of fitted logistic regression model 18:02 How to report results
Thank you for this great video!! At the beginning you mentioned that the EV can be on an ordinal scale too. However, is that totally correct? Because is it correct to interpret the beta coefficient (e.g., a unit increase in X increases the log odds by beta units) when the one-unit steps at the EV are not the same for each EV category (as common for ordinal-scaled variables)? For me, this makes only sense if you treat it as a dummy-coded variable (with one reference category) as you did in the video, but not, if you just put it in as a "normal" covariate (because then it is treated as continous).