I have a continuous instrument that is strongly associated with the exposure. I have also confirmed statistically that my selected instrument is strong and relevant. My concern is with compliers. I understand that under the monotonicity assumption, all four subgroups (i.e., compliers, defiers, always-takers, never-takers) cannot be distinguished since we are dealing with counterfactual scenarios. However, would a strong association between the instrument and exposure be interpreted as my sample being composed mostly of compliers?
Asked
Active
Viewed 97 times
