Table 6.
The first two columns show the pooled estimated coefficients and standard errors (SE) from the linear regressions on Y2 for Models 1 and 2 or Y4 for Model 3 (n = 300) using the full datasets (FD). The remaining columns show the average relative differences (%) as compared to those from FD for the complete-case analysis (CC), log-normal imputation (LMlog), predictive mean matching (PMM), and symmetric (QRTs) and asymmetric (QRTa) transformed quantile regression imputation models. The latter were fitted with either (1) known or (2) unknown λp.
| FD | CC | LMlog | PMM | QRTs (1) | QRTa (1) | QRTs (2) | QRTa (2) | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Est. | SE | Est. | SE | Est. | SE | Est. | SE | Est. | SE | Est. | SE | Est. | SE | Est. | SE | |
| Model 1 | ||||||||||||||||
| θ0 | 0.2 | 0.1 | 13.6 | 23.6 | −5.0 | 9.9 | −1.3 | 3.7 | 0.6 | 3.6 | 0.7 | 3.8 | −3.6 | 5.2 | 0.3 | 4.3 |
| θ1 | 1.9 | 0.2 | −1.3 | 5.8 | −10.5 | 59.3 | −0.2 | 3.4 | 0.3 | 4.0 | −0.7 | 4.9 | −1.3 | 7.0 | −1.1 | 5.4 |
| θ2 | −0.6 | 4.9 · 10−2 | −0.2 | 14.2 | −1.9 | 16.6 | −0.1 | 3.5 | 0.1 | 3.4 | 0.3 | 3.7 | −0.7 | 5.1 | 0.0 | 4.1 |
| θ3 | 0.3 | 2.9 · 10−2 | −1.8 | 9.9 | 4.8 | 17.6 | 0.2 | 2.3 | −0.1 | 2.5 | 1.0 | 2.9 | 0.2 | 3.6 | 0.8 | 3.2 |
| Model 2 | ||||||||||||||||
| θ0 | 1.8 · 10−2 | 0.1 | −39.7 | 25.1 | −223.0 | 3.7 | −41.9 | 1.3 | −15.2 | 0.7 | −38.4 | 0.9 | −19.1 | 0.8 | −39.9 | 0.9 |
| θ1 | 0.5 | 0.1 | −9.1 | 5.0 | −58.1 | 33.4 | −4.6 | 5.4 | −0.9 | 5.7 | −3.2 | 5.9 | −1.3 | 5.6 | −3.4 | 5.8 |
| θ2 | −0.5 | 0.1 | −1.2 | 15.5 | −7.0 | 4.6 | −0.7 | 1.1 | −0.2 | 0.8 | −0.5 | 0.9 | −0.3 | 0.8 | −0.5 | 0.9 |
| θ3 | 0.5 | 2.9 · 10−2 | 0.6 | 10.1 | 3.8 | 5.4 | 0.5 | 0.9 | 0.1 | 0.8 | 0.4 | 0.8 | 0.1 | 0.9 | 0.4 | 0.8 |
| Model 3 | ||||||||||||||||
| θ0 | 1.1 | 0.1 | 16.7 | −5.0 | −2.8 | −22.5 | −0.2 | 52.8 | −0.0 | 28.8 | 0.2 | 30.7 | 0.3 | 31.2 | 0.2 | 30.8 |
| θ1 | −0.1 | 0.1 | −29.0 | −3.6 | −58.8 | −46.2 | −10.7 | 69.1 | −4.0 | 38.2 | −0.1 | 40.2 | 1.3 | 40.8 | 0.1 | 40.4 |