Given the high number of explanatory variables, model selection was performed following two steps.
With a growing number of variables model selection becomes computationally demanding and quickly exceeds feasibility, thereby making the method difficult to adapt to high-dimensional data.
It also focuses on statistical issues of scale, techniques for defining resource use and availability, pooling observations, independence of relocations, and variable and model selection and how these factors affect inference in resource selection studies.
The bootstrapping technique used for variable and model selection was proposed in the 1990's [ 42] and has been discussed and applied in different studies [ 43, 44].
To avoid having to fit all possible models, we used a selection procedure that compared models with major differences in their predictor variables (see model selection, below).
The MFP method also scaled and centered variables in model selection process to improve numerical stability and to provide a model intercept that was easier to interpret.
Predictor variables and model selection are described in detail in the supplementary material, but we will briefly mention important predictors here.
Among these variables, the model selection procedure identified agricultural yield as the most important factor influencing total haemosporidian prevalence (Table 2 and Fig. 2A).
A simulation study is carried out to demonstrate the merit of designs of variable resolution in model selection.
For example, the latent variable Gaussian graphical model selection in [1], the quadratic discriminant analysis model in [2] and the robust principal component analysis model with noisy and incomplete data in [3, 4], and so on.
As genotyping technology rapidly advances, and genome-wide association studies become increasingly important in the field, the importance of variable selection in model selection is highlighted.
