Abstract:
|
RTI developed the Cyclical Tree-Based Hot Deck (CTBHD) imputation method, which implements a regression tree to construct imputation cells to form imputation classes and uses weighted sequential hot deck imputation to select the donor. The imputation process is repeated in cycles with the goal to attain a better result; that is, to have the imputed values converge to stable values. The base cycle follows a single imputation process, where only the complete response variables and any previously imputed variables are used in the tree models. In any subsequent cycles, variables are re-imputed by updating the imputation classes, which are created by using all the variables in the set. However, the properties of cycling are not known and have not been studied. Here we evaluate whether CTBHD gets an advantage from cycling the imputation using data from the Residential Energy Consumption Survey (RECS). We will investigate the ideal number of cycles needed to attain the optimal imputed results. The results from multiple cycles of imputation will be compared to the single hot deck imputation results.
|