2.4. Statistical analyses
Before conducting statistical analyses, the Kolmogorov-Smirnov and Levene’s tests were employed to assess data normality and homogeneity of variance, respectively.
To analyze the habitat selection preference of D. splendidum across different sex and age groups, Pearson correlation analysis was conducted to assess autocorrelation among numerical ecological factors, Kendall correlation analysis was conducted to assess autocorrelation among classified ecological factors, and Spearman correlation analysis was conducted to assess autocorrelation between numerical and classified ecological factors. Factors exhibiting high autocorrelation (|r| > 0.8; Gardiner et al., 2015; Chiaverini et al., 2023) were subsequently excluded. Perch height, reflecting species-specific preferences, was omitted from habitat selection preference analysis. Subsequently, a random forest model was constructed using these numerical and classified factors, and trend charts were generated for the Mean Decrease Gini index (Zhang et al., 2020; R Development Core Team, 2018). Based on the differential importance of predictor variables indicated by the mean decrease Gini index, partial dependence plots were generated for various habitat factors.
To compare differences in habitat preference factors among different sex and age groups, a general linear model (GLM) was initially employed to quantify the effects of sex, age, and their interactions on habitat selection (Delaney et al., 2016). Subsequently, simple effect analysis was conducted to evaluate the influence of ecological factor interactions, excluding non-significant interactions. For numerical factors conforming to a normal distribution, differences in habitat selection between different sex and age groups were assessed using independent sample t -tests. In cases where numerical factors did not follow a normal distribution, differences in habitat selection among populations were assessed using the Mann-Whitney U test (Li et al., 2022). Chi-square tests were performed to analyze disparities in classified factors, such as slope direction and vegetation type, regarding sex and age preferences (Shilereyo et al., 2021). When significant differences were observed (P < 0.05), Vanderploeg & Scavia’s selection index was employed to elucidate population-specific habitat preferences.
All statistical analyses were conducted using R Studio (v23.3.1, Racine et al., 2012) and SPSS software (v22.0).