Graphical Criteria of Recoverability under Not Missing at Random Case
Date:
In this study, I suggested a causal graphical approach using m-DAG to check the possible bias in either in intercept and slope under the list-wise deletion situation. By doing so, this paper proposes general criteria for identifying which parameters are recoverable without bias under NMAR. Also, using m-DAGs with distinct notations for conditioning and controlling, this paper clarifies mechanisms behind bias in NMAR models. This approach allows researchers to assess recoverability of means and slopes (e.g., treatment effects), even under listwise deletion.