Table 2.
Average Bias, RMSE, and Correlation Values for NMAR.
| Method | Test length | Missing proportion (%) | Bias | RMSE | Correlation |
|---|---|---|---|---|---|
| FIML | 20 | 30 | 0.000 | 0.348 | .911 |
| MICE-RFI2 | 20 | 30 | 0.000 | 0.371 | .898 |
| MICE-CART | 20 | 30 | 0.000 | 0.373 | .897 |
| MICE-CART2 | 20 | 30 | 0.000 | 0.375 | .896 |
| MICE-RFI | 20 | 30 | 0.000 | 0.379 | .893 |
| Zero replacement | 20 | 30 | 0.000 | 0.466 | .834 |
| FIML | 40 | 30 | 0.000 | 0.291 | .947 |
| MICE-CART | 40 | 30 | 0.000 | 0.312 | .939 |
| MICE-RFI2 | 40 | 30 | 0.000 | 0.318 | .937 |
| MICE-CART2 | 40 | 30 | 0.000 | 0.319 | .936 |
| MICE-RFI | 40 | 30 | 0.000 | 0.320 | .936 |
| Zero replacement | 40 | 30 | 0.000 | 0.395 | .901 |
| FIML | 60 | 30 | 0.000 | 0.254 | .962 |
| MICE-CART | 60 | 30 | 0.000 | 0.270 | .957 |
| MICE-CART2 | 60 | 30 | 0.000 | 0.281 | .954 |
| MICE-RFI2 | 60 | 30 | 0.000 | 0.281 | .954 |
| MICE-RFI | 60 | 30 | 0.000 | 0.282 | .954 |
| Zero replacement | 60 | 30 | 0.000 | 0.350 | .928 |
| FIML | 20 | 40 | 0.000 | 0.433 | .860 |
| MICE-CART2 | 20 | 40 | 0.000 | 0.456 | .842 |
| MICE-RFI2 | 20 | 40 | 0.000 | 0.460 | .839 |
| MICE-CART | 20 | 40 | 0.000 | 0.463 | .839 |
| MICE-RFI | 20 | 40 | 0.000 | 0.473 | .827 |
| Zero replacement | 20 | 40 | 0.000 | 0.503 | .804 |
| FIML | 40 | 40 | 0.000 | 0.370 | .914 |
| MICE-CART | 40 | 40 | 0.000 | 0.386 | .905 |
| MICE-CART2 | 40 | 40 | 0.000 | 0.393 | .901 |
| MICE-RFI2 | 40 | 40 | 0.000 | 0.401 | .898 |
| MICE-RFI | 40 | 40 | 0.000 | 0.409 | .893 |
| Zero replacement | 40 | 40 | 0.000 | 0.434 | .879 |
| FIML | 60 | 40 | 0.000 | 0.331 | .936 |
| MICE-CART | 60 | 40 | 0.000 | 0.346 | .929 |
| MICE-CART2 | 60 | 40 | 0.000 | 0.352 | .926 |
| MICE-RFI2 | 60 | 40 | 0.000 | 0.364 | .922 |
| MICE-RFI | 60 | 40 | 0.000 | 0.370 | .918 |
| Zero replacement | 60 | 40 | 0.000 | 0.386 | .911 |
| FIML | 20 | 70 | 0.000 | 0.680 | .618 |
| MICE-CART | 20 | 70 | −0.004 | 0.708 | .585 |
| MICE-RFI2 | 20 | 70 | 0.000 | 0.733 | .491 |
| MICE-RFI | 20 | 70 | 0.000 | 0.738 | .487 |
| MICE-CART2 | 20 | 70 | 0.000 | 0.754 | .485 |
| Zero replacement | 20 | 70 | 0.000 | 0.777 | .463 |
| FIML | 40 | 70 | 0.000 | 0.635 | .728 |
| MICE-CART | 40 | 70 | −0.010 | 0.652 | .708 |
| MICE-RFI | 40 | 70 | −0.001 | 0.683 | .660 |
| Zero replacement | 40 | 70 | 0.000 | 0.687 | .665 |
| MICE-RFI2 | 40 | 70 | −0.001 | 0.706 | .634 |
| MICE-CART2 | 40 | 70 | 0.001 | 0.727 | .606 |
| FIML | 60 | 70 | 0.001 | 0.595 | .779 |
| MICE-CART | 60 | 70 | −0.011 | 0.601 | .770 |
| Zero replacement | 60 | 70 | 0.000 | 0.635 | .739 |
| MICE-RFI | 60 | 70 | −0.001 | 0.643 | .727 |
| MICE-RFI2 | 60 | 70 | −0.001 | 0.676 | .698 |
| MICE-CART2 | 60 | 70 | 0.001 | 0.682 | .683 |
Note. RMSE = root mean square error; NMAR = not missing at random; FIML = full-information maximum likelihood; MICE-CART = multiple imputation with chain equations utilizing classification and regression trees; MICE-RFI = multiple imputation with chain equations utilizing random forest imputation.