# Interpret Coefficeint as change in Percentage-point or Percentage

Having carried out the regression below, I'm struggling to determine what the correct interpretation of the predictor variable would be.

Given that the dependent variable is binary, where 1=receiving a medicine and 0= not receiving a medicine. And the predictor variable is a measure of distance measured in km from a hospital.

Would the correct interpretation be that a 1km increase in distance from the hospital leads to a 2.26 percentage-point increase in the probability of receiving the medicine?

OR, a 1km increase in distance from the hospital leads to a 2.26% increase in the probability of receiving the medicine?

I'm trying to understand the difference between the two interpretations and which would be correct wrt to the regression? • Did you perform normal regression or logistic regression? If the output is binary, a normal regression is not the correct tool to use. – Davide ND Jan 7 at 12:56
• The regression was in fact the First stage of a Two Stage Least-Square computation. Even in that case would I have had to use another type of regression? – SoniaG Jan 7 at 13:10
• I dont know what you mean by two stage computation. However, if the predicted value is binary then you should you a Logit or a Probit regression - which predict the probabilty instead of the value. – Davide ND Jan 7 at 13:29
• If you're using 2SLS, what are your true variables? You can do 2SLS with binary outcomes, but I believe the predictors must also be binary. Were you told to use 2SLS here? Also, 0.226261 would be 22.61%. – Todd Burus Jan 7 at 13:31