Fit indices, chi-square difference tests, residuals and factor loadings were considered for model comparison.
The following discusses maximum likelihood estimation of the change point models, model comparison, the estimation of random effects, and the assessment of residuals.
A statistical model comparison [ 21] was applied to test whether the differences between the cross-validated residuals from models with different number of extracted factors were significant.
For a given gene, model comparison between σ e (i 1 ) 2 ≠ σ e (i 2 ) 2 and σ e (i 1 ) 2 = σ e (i 2 ) 2 hypotheses simplifies to conditions π i ≠ 0.5 (within-gene heterogeneous residual variances for all genes; Model HE) and π i = 0.5 (homogeneous residual variance for the ith gene, within-gene heterogeneous residual variances for the remaining genes; Model HO i ).
After examination of parameter estimates, fit indexes, and residuals, model comparisons and model modifications to the original hypothesized model were a priori planned to identify any possible better fitting and more parsimonious models [ 21].
When multiple inputs were changed simultaneously, the correlation between residuals diminished more (r ranging from 0.77 to 0.85, Table 3).> -wrap-foot> Upper diagonal shows average correlation of residual outcome for each chromosome for 15 pairwise model comparisons.
Correlations between the chromosome-specific residual sets were computed for all 15 pairs of model comparisons, as well as for the corresponding predicted phenotypic value sets.
Although original and updated BLUPs were highly correlated as input values (r =0.99), their small differences resulted in QTL model residuals with much greater differences, reflected in correlations of 0.80 to 0.91 between residual outcomes differing only for original vs. updated BLUPs (Model comparisons 1 vs 3, 2 vs. 4, and 5 vs. 6; Table 3).
Make price and model comparisons.
