Table 2. Comparison between models to predict the left ventricular (LV) end-diastolic dimension and LV mass with or without fat-free mass (FFM) as a covariate.
| FFM as a covariate | –2LL | Number of parameters | Change in –2LL for χ 2 test (df, N) | ||
|---|---|---|---|---|---|
| LV end-diastolic dimension | |||||
| linear models | model-1 | – | 603.628 | 5 | |
| model-3 | + | 582.714 | 6 | 20.914 * (df = 1, N = 131) | |
| quadratic models | model-2 | – | 599.077 | 6 | |
| model-4 | + | 581.285 | 7 | 17.792 * (df = 1, N = 131) | |
| LV mass | |||||
| linear models | model-1 | – | 1082.599 | 5 | |
| model-3 | + | 1062.175 | 6 | 20.424 * (df = 1, N = 131) | |
| quadratic models | model-2 | – | 1070.193 | 6 | |
| model-4 | + | 1054.456 | 7 | 15.737 * (df = 1, N = 131) | |
* p < 0.01,
–2LL: minus twice the log-likelihood, N: total number of measurements.
Chi-square (χ 2) statistics were calculated by subtracting the –2LL of model-3 from that of model-1 and by subtracting the –2LL of model-4 from that of model-2 for linear and quadratic model comparisons, respectively. Degrees of freedom (df) for χ 2 tests were calculated by subtracting the number of parameters in model-1 from that in model-3 and by subtracting the number of parameters in model-2 from that in model-4 for comparison of quadratic models comparisons.