Table 2.
Model fit for response mappings and ALDVMMs
| FACT-B total/subscales | Number of components | Log likelihood | AIC | BIC | Mean error | Mean absolute error | RMSE | Proportion of predicted observations at full health (%) | Proportion of predicted observations below 0 (%) |
|---|---|---|---|---|---|---|---|---|---|
| Response mapping | |||||||||
| Total scores | – | −29,869.95 | 59,809.91 | 60,068.53 | 0.0117 | 0.1341 | 0.1824 | 21.7 | 2.8 |
| Subscales | – | −27,530.01 | 55,170.02 | 55,576.43 | 0.0092 | 0.1216 | 0.1682 | 29.6 | 1.89 |
| ALDVMMs | |||||||||
| Total FACT-B | 1 | −735.2737 | 1,480.547 | 1,517.493 | 0.0018 | 0.1330 | 0.1829 | 34.3 | 0.2 |
| 2 | 1,217.156 | −2,406.312 | −2,302.863 | 0.0006 | 0.1324 | 0.1815 | 25.2 | 1.6 | |
| 3 | 1,395.884 | −2,745.768 | −2,575.818 | −0.000007 | 0.1318 | 0.1815 | 27.4 | 1.5 | |
| 4 | 1,724.382 | −3,384.763 | −3,148.310 | 0.0001 | 0.1332 | 0.1818 | 27.4 | 1.5 | |
| Subscales | 1 | 435.7411 | −853.4822 | −786.9798 | −0.0002 | 0.1207 | 0.1656 | 33.9 | 0.1 |
| 2 | 2,476.873 | −4,901.747 | −4,709.629 | 0.0005 | 0.1199 | 0.1654 | 25.9 | 1.6 | |
| 3 | 2,867.129 | −5,648.258 | −5,330.524 | −0.0004 | 0.1155 | 0.1631 | 27.6 | 1.4 | |
| 4 | 3,075.795 | −6,031.590 | −5,588.240 | −0.00009 | 0.1158 | 0.1633 | 27.0 | 1.5 | |
Note. Proportion of observations at full health = 27.6%, proportion in a state worse than death = 1.4%. Note that AIC and BIC are not comparable between response mapping and ALDVMMs.
AIC, Akaike information criterion; ALDVMM, adjusted limited dependent variable mixture model; BIC, Bayesian information criterion; FACT-B, Functional Assessment of Cancer Therapy—Breast Cancer; RMSE, root mean square error.