Statistical analyses
We calculated the “thermal breadth (°C)” by maximum - minimum habitat temperature (°C). We obtained n = 306 data records of thermal breadth (°C).
All statistics and graphics were performed using R ver. 4.0.2 (R Core Team, 2020). We set the significant level, α = 0.05. We performed generalized linear models (GLMs) to predict the maximum and minimum temperature and thermal breadth using “glm” function with the degree of latitude N of the collected points, the elevation (above sea level, m), and their mean body length as the body size index. We used the Gaussian distribution as the error distribution for the GLM. We performed the GLMs without the random factor due to too many data source categories from 967 independent reports for the database. Before the GLM analysis, we calculated the variance inflation factor (VIF) to check for co-linearity of among factors. The maximum VIF was 10.2 for the explanatory factors of GLM so we excluded the maximum elevation for all GLMs because of significantly correlated with minimum elevation (Pearson’s coefficient by “cor.test” function, r = 0.692, P < 0.0001). After the exclusion, the maximum VIF was 1.34 indicating that co-linearity among the factors would not significantly influence the results of GLMs. We used simple GLM without random factors, because the database does not have the suitable random factors for the analysis.
For categorical data, including voltinism, functional feeding guilds, and taxonomic Order, we performed a three-way analysis of variance (ANOVA) and evaluated interactions using “aov” function. Also, we simply used the ANOVA without the random factor due to 967 independent reports for the database. When the ANOVA was significant, we performed the post-hoc test using Tukey multiple comparisons with “TukeyHSD” function.