Today, someone asked me for feedback on his research. I just went through it; he has regressed a binary outcome applying an OLS model.
I tried to explain why he needs to use a logit model: 1) SEs and confidence intervals will be invalid..
#EconTwitter #AcademicChatter

since they are by default calculated assuming normality for the outcome (conditional on covariates)
2) variance is a function of the mean in binary data and is not constant as the mean changes. This violates the assumption that the variance of the residual errors is constant

3) it is entirely possible to have a fitted regression which gives predicted values for some individuals which are outside of the (0,1) range or probabilities
4) just use logit, man

He stared at me like I was speaking French. I decided to get into his do file and help him out. Because when I needed it, @fnajeeb42 was the one to spend day and night helping me out with my thesis while not making me feel as dumb as I was at the time

She’s now a PhD student and still the smartest woman I know. Part of the reason I know enough econometrics to help another human out. It is in her name that I’m doing this. Help out your dumb friends people, it could be sadqa e jaariya for you

#Econometrics #statstwitter