I am supposed to run a regression of Medical Expenditure on several explanatory variables like income, no of illnesses, age and also some dummy variables like gender differentiating dummy, dummy for health insurance cover etc. The simple linear regression thus obtained turns out to have low t and R2 values. I used some other functional forms like log transformation for income and medical expenditure. I even tried dropping some variables but the insignificance and low R2 persist. I dropped the intercept and the regression turned out to have high R2 (almost .95 in comparison to a .15 in the first model). How do I interpret this? I know dropping the intercept has some implications and we need to factor in presence of Dummy Variables as well since without an intercept, there'd be no base case for the Dummy Variables. What is wrong with the entire interpretation? Given the possibility of dropping a variable, how do I interpret the dummy variables (there are 4 in total).

  • $\begingroup$ See this post. $\endgroup$ – Dimitriy V. Masterov Nov 5 '18 at 22:54
  • 1
    $\begingroup$ What does the distribution of Medical Expenditure look like? Do you have any records for which Medical Expenditure is equal to zero and if yes, how many? $\endgroup$ – Isabella Ghement Nov 5 '18 at 23:11