**2. Materials and Methods**

Based on the previous discussion of diversified public health impacts in the Arctic, the authors applied multiple regression analysis to reveal the variables *Xn* that affect the incidence rates of health disorders *Yn*. The six-stage algorithm was employed (Figure 2).

**Figure 2.** Study flow algorithm. Source: authors' development.

The study started with a selection of *Xn* regressors to be considered for inclusion in the model and development of the regression model (Stage 1). To avoid redundancy, variance inflationary factor (VIF) was computed for each *Xn* at Stage 2. Based on the criteria developed by Snee [57] and further applied by Kutner et al. [58], Montgomery et al. [59], and Ermakov et al. [60], that VIF values should be less than 5, those *Xn* for which *VIF* > 5 were excluded from the model. At Stage 3, a best-subsets regression was performed with the remaining *Xn* for all models. To finalize the collinearity test, the parameters of adjusted *R*<sup>2</sup> [61,62] and Mallows' *Cp* statistic [63–66] were computed for each subset. The subsets with *Cp* > (*k* + 1) were eliminated; the study proceeded with those "best" subsets for which relative *Cp* were the lowest and/or adjusted *R*<sup>2</sup> were high. At Stage 4, multiple regression analysis of the models chosen was performed across *Yn* regressands and territories. The revealed correlations allowed us to categorize the territories based on several parameters (Stage 5) and discover the effects of *Xn* regressors on *Yn* regressands (Stage 6).
