Suppose we'd like to see some interactions with certain diseases and SES. So, we select all the individuals with the diseases from the entire population data, say 5%. Then we select other 5% from the rest of the population with the stratified random sampling by age-sex-income. Then, we'll see time to death through survival analyses by including both diseases, SES, and other covariates. Here are my questions.
-Is it possible to assume that the 5% of the population without the disease is a representative sample of the rest 90%?
-If we don't have the weight variable, what procedures can be used to test that distributions of predictors?
-are the estimates from Cox proportional analyses biased in this case?
