Table 1. The eight tested models, model outcome in terms of the temporal pattern of survival differences between larval types.
| Model (i)* | AIC | Δi | Wi | Temporal pattern† |
|---|---|---|---|---|
| D warning-coloured: identical linear cryptic: non-linear | 1,449.21 | 0 | 0.36 | Both signals: W (5 June) C (27 July) W |
| E warning-coloured: linear diff.intercept cryptic: non-linear | 1,449.28 | 0.07 | 0.35 | Small signal: W (3 June) C (30 July) WLarge signal: W (8 June) C (24 July) W |
| F warning-coloured: different linear cryptic: non-linear | 1,450.09 | 0.88 | 0.23 | Small signal: W (31 July) C (28 July) WLarge signal: W (10 June) C (26 July) W |
| H all different, non-linear | 1,452.76 | 3.55 | 0.06 | Small signal: W (28 May) C (24 July) WLarge signal: W (7 June) C (22 July) W |
| G all identical, non-linear | 1,459.91 | 10.70 | <0.01 | No difference between larval types |
| A all identical, linear | 1,466.85 | 17.64 | <0.01 | Survival of all types increases with time |
| B all linear, warning-c diff. from cryptic | 1,470.23 | 21.02 | <0.01 | Survival of all types increases with time |
| C all different and linear | 1,471.33 | 22.12 | <0.01 | Survival of all types increases with time |
AIC, Akaike Information Criteria.
Models are listed in order of increasing AIC values (AIC, score differences Δi and Akaike weights Wi).
*The model description uses ‘identical’ to denote that parameter values were shared between larval types, and ‘different’ to denote that parameters were estimated separately. ‘Nonlinear’ refers to a quadratic form (survival=a+bd+c2), where d=date (from the beginning of the year), and a, b and c are parameters. ‘Linear’ refers to survival=a+bd, and ‘diff.intercept’ to a linear model in which two larval types share the same slope b but the intercept a is estimated separately for each type.
†The column ‘temporal pattern’ summarizes the outcome by giving the date-specific winning type. Thus ‘W (5 June) C (27 July) W’ means that the model predicts the warning-coloured type to have higher survival than the cryptic type before 5 June and after 27 July but not between those dates. This is given separately for the different warning-colour signals where the model differentiates between them.