Table 5. Comparison of multilocus coat pheomelanin intensity predictive models.
| Variables | β ± se | t-value | P>|t| | PRE | Adj. R2 | ln(likelihood) | |
|---|---|---|---|---|---|---|---|
| A. | Intercept | 1.012 ± 0.049 | 20.831 | <2.2x10-16 | - | 0.7300 | -2,795.30 |
| CFA2 | 0.915 ± 0.026 | 35.074 | <2.2x10-16 | 0.365 | |||
| CFA15 | 0.191 ± 0.026 | 7.225 | <2.2x10-16 | 0.024 | |||
| CFA18 | 0.272 ± 0.056 | 4.85 | <2.2x10-16 | 0.011 | |||
| CFA20 | 1.038 ± 0.026 | 39.262 | <2.2x10-16 | 0.419 | |||
| CFA21 | 0.215 ± 0.027 | 0.027 | <2.2x10-16 | 0.029 | |||
| B. | Intercept | 1.074 ± 0.043 | 25.088 | <2.2x10-16 | - | 0.7324 | -2,785.92 |
| CFA2 | 0.920 ± 0.026 | 35.666 | <2.2x10-16 | 0.373 | |||
| CFA15_2 | 0.286 ± 0.039 | 7.256 | <2.2x10-16 | 0.024 | |||
| CFA18_red_dom | 0.405 ± 0.074 | 5.444 | <2.2x10-16 | 0.014 | |||
| CFA20 | 1.037 ± 0.026 | 39.453 | <2.2x10-16 | 0.421 | |||
| CFA21_red_dom | 0.355 ± 0.040 | 8.904 | <2.2x10-16 | 0.036 | |||
| C. | Intercept | 1.606 ± 0.062 | 25.834 | <2.2x10-16 | - | 0.5394 | -3375.46 |
| CFA15_2 | 0.053 ± 0.096 | 0.550 | 5.82 x 10−1 | 0.000 | |||
| CFA15_2 x CFA20 | 0.374 ± 0.063 | 5.956 | <2.2x10-16 | 0.016 | |||
| CFA20 | 1.290 ± 0.043 | 29.844 | <2.2x10-16 | 0.294 | |||
| D. | Intercept | 1.095 ± 0.054 | 20.250 | <2.2x10-16 | - | 0.7353 | -2772.11 |
| CFA2 | 0.908 ± 0.026 | 35.087 | <2.2x10-16 | 0.366 | |||
| CFA15_2 | 0.167 ± 0.081 | 2.050 | 4.1 x 10−2 | 0.002 | |||
| CFA15_2 x CFA20 | 0.161 ± 0.049 | 3.291 | 1.0 x 10−3 | 0.005 | |||
| CFA15_2 x CFA21_red_dom | -0.139 ± 0.079 | -1.752 | 8.0 x 10−2 | 0.001 | |||
| CFA18_red_dom | 1.225 ± 0.217 | 5.65 | <2.2x10-16 | 0.015 | |||
| CFA18_red_dom: CFA20 | -0.381 ± 0.112 | -3.400 | 1.0 x 10−3 | 0.005 | |||
| CFA18_red_dom: CFA21_red_dom | -0.308 ± 0.174 | -1.772 | 7.7 x 10−2 | 0.001 | |||
| CFA20 | 0.985 ± 0.034 | 28.85 | <2.2x10-16 | 0.281 | |||
| CFA21_red_dom | 0.436 ± 0.055 | 7.944 | <2.2x10-16 | 0.029 | |||
| E. | Intercept | 1.134 ± 0.051 | 22.195 | <2.2x10-16 | - | 0.7346 | -2,775.53 |
| CFA2 | 0.908 ± 0.026 | 35.043 | <2.2x10-16 | 0.365 | |||
| CFA15_2 | 0.102 ± 0.073 | 1.387 | 1.67x10-1 | 0.001 | |||
| CFA15_2 x CFA20 | 0.148 ± 0.048 | 3.061 | 2.0x10-3 | 0.004 | |||
| CFA18_red_dom | 1.017 ± 0.185 | 5.496 | <2.2x10-16 | 0.014 | |||
| CFA18_red_dom x CFA20 | -0.406 ± 0.112 | -3.640 | <2.2x10-16 | 0.006 | |||
| CFA20 | 0.992± 0.034 | 29.141 | <2.2x10-16 | 0.285 |
Coefficients, coefficient standard error, t score values, t test p-values, and PRE for the y-intercept and each of the independent variables in the best fit linear model incorporating non-additivity and pairwise epistasis. Section A. shows the base model that assumes perfect additivity at each locus and no interactions between loci. Section B. shows the best fit model incorporating dominance at all five loci. Section C. shows a model consisting of only the two previously reported loci (CFA15 and CFA20) using their best dominance encoding, and their pairwise interaction (CFA15_2 x CFA20). Section D. shows the best fit model incorporating both the dominance terms in model B. and two pairwise epistasis terms: CFA15_2 x CFA20 and CFA18_red_dom x CFA20. Section E. shows a reduced version of model D. that only includes terms that explained > 0.1% of variance (PRE > 1 x 10−3) in model D. and shows similar performance.