TABLE 3.
Model selection analyses for TuMV and CMV percentages and total numbers of infected seedsg
| Model structurea | R2cb | Log likelihood | AICc | Δid | ωie |
|---|---|---|---|---|---|
| ST (%) | |||||
| G, −4.55 + 10.30 · VAI (53) + 3.75 · SM (41) − 1.81 · IWi/IWm (2) − 4.19 · V (1) + 2.10 · VAR (1) + 4.80 · G0i/G0m (1) − 2.88 · G24i/G24m (1) | 0.91* | −226.43 | 472.86 | 2 | 0.48 |
| T, −13.59 + 1.49 · VAI (53) + 4.58 · SM (24) − 1.98 · V (14) − 1.49 · IWi/IWm (6) + 12.62 · G0i/G0m (2) − 0.88 · G24i/G24m (1) | 0.99* | −101.73 | 219.45 | 8 | 0.20 |
| C, −1.13 + 1.09 · VAI (37) + 3.19 · SM (37) − 0.75 · G48i/G48m (18) + 0.60 · IWi/IWm (8) | 0.71* | −85.14 | 182.27 | 5 | 0.33 |
| IS (log n)f | |||||
| G, 0.95 + 0.06 · ST (41) + 0.24 · SNi (23) − 0.59 · V (20) + 0.18 · VAI (10) − 0.42 · RWi/RWm (4) − 0.39 · G48i/G48m (2) | 0.78* | −74.49 | 168.98 | 6 | 0.25 |
| T, −2.59 + 0.05 · ST (45) + 0.97SNi (20) − 1.18 · V (15) + 1.17 · SM (7) + 0.21 · VAI (7) − 1.35 · IWi/IWm (4) − 0.48 · G48i/G48m (2) | 0.81* | −32.01 | 86.23 | 9 | 0.18 |
| C, 0.45 + 0.12 · ST (25) − 0.36 · V (24) + 1.90 · 10−5 · SNi (20) + 1.90 · SM (13) 0.11 · VAI (10) − 1.22 · RWi/RWm (4) + 1.08 · IWi/IWm (4) | 0.78* | −39.37 | 96.74 | 15 | 0.13 |
The relative importance (percent) of each estimator variable is shown in parentheses. G, global model; T, TuMV-specific model; C, CMV-specific model.
Conditional correlation coefficient. Asterisks indicate significant correlations (P < 0.01).
AIC, Akaike’s information criterion.
Number of models closely competing with the best-ranked model (Δi of <2 out of 511 for ST and 2,047 for IS models tested). Δi is the difference between the AIC of a given model and that of the best-ranked model and quantifies how models compete (for the best-ranked model, Δi = 0; for substantial empirical support, Δi = 1 to 2; for considerably less support, Δi = 2 to 7; for no support, Δi > 10) (68).
AIC model weight as ωi = exp(−0.5Δi)/Σexp(−0.5Δi). The larger the ω value, the greater the likelihood of the model relative to the competing models. The maximum ωi is 1.
The number of infected seeds (IS) was normalized using a logarithmic transformation, and the resulting values were used for model construction.
Model structures for ST included the effect of virus infection on rosette (RWi/RWm) and inflorescence (IWi/IWm) weights and on short-term (G0i/G0m), medium-term (G24i/G24m), and long-term (G48i/G48m) seed survival; virulence (V); virus within-host speed of movement (SM); and virus accumulation in rosette (VAR) and inflorescence (VAI) leaves. Model structures for IS also included ST and the total log number of seeds produced per plant (SN). Best-ranked models are shown.