Which statement is incorrect? Explain all the options.
1) It often makes sense to omit variables from a regression to make it causal, even if they would improve the R2 for the regression.
2) In some contexts, coarsening strata can reduce bias when using stratification methods (relative to not coarsening).
3) Experiments are usually the preferred method for doing causal inference when they are possible.
4) In theory, bias in an ATE estimate will always decrease as you block more back-door paths.