In this study, we consider linear social interaction models under incomplete information that allow for missing outcome data due to sample selection. For model estimation, assuming that each individual forms his/her belief about the other members' outcomes based on heterogeneous rational...