Table 1.

Statistical summaries of the best binomial models of thermal tolerance. (The first row for each analysis contains p-values for each of the included predictors (values in bold are significant, α < 0.05), and subsequent rows contain corresponding odds ratios for each term in the model (with 95% confidence intervals). Predictors left blank were excluded from the final model. Odds ratios were produced by exponentiating the model β coefficients and indicate the odds of survival relative to the reference level for each variable. Quasi-complete separation required some model coefficients to be derived by exact conditional logistic regression using the elrm package in R (*). Although the interaction between origin and treatment was initially included in the full model for all analyses, it was not included in any of the best models and is thus not shown here.)

analysis	origina	treatmentb	temperature	origin × temperaturec
juvenile one month	p = 0.002 highprot = 1.13 (0.42, 3.03) lowexp = 0.17 (0.05, 0.50) lowprot = 0.69 (0.25, 1.88)	p < 0.001 exp = 0.73 (0.24, 2.22) field = 2.63 (0.90, 8.09) prot = 10.82 (3.84, 34.00)	p < 0.001 temp = 0.09 (0.05, 0.14)
juvenile six months + common garden			p < 0.001 temp* = 0.75 (0.00, 0.85)
adult one month			p < 0.001 temp = 0.04 (0.01, 0.12)	p < 0.001 temp × prot = 1.12 (1.05, 1.24)

^aReference = high exposed.

^bReference = common garden.

^cReference = exposed × temperature (adults only).