Missing data in IV
What should you do with missing data in an instrumental variables analysis? For my dissertation, I’m using instrumental variables to examine the impact of family income on children’s cognitive skills. Having been trained as a psychologlist, there is an emphasis on handling missing data appropritately – Full Information Maximum Likelihood (FIML) and Multiple Imputation (MI) are popular techniques. However, in the causal inference courses I’ve taken, there seems to be less of a focus on how to handle missing data.