Model 1 showed improved goodness-of-fit and somewhat improved discrimination; however, the discriminative ability was not significantly enhanced from the PRESERVE score.
The comparison of predicted and observed values of proportion of missing data from the BRT model in figure 3B confirmed that this model provides improved goodness-of-fit.
19 20 A forward variable-selection approach was used and variables retained if they improved goodness-of-fit, based on the Akaike information criterion.
The plots refer to the underlying models which were used to estimate the borders of the PIs and show a slightly improved goodness-of-fit for quantile boosting.
Significantly improved goodness-of-fit statistics for the one-factor model and/or the model controlled by the Common Latent Factor would signify the problem of common method bias.
Introducing third component in the envelope of fitting curves did not improve goodness of fitting significantly (the bound positron-electron states were not proper for studied Se-based ChG in full agreement with previous results [17, 18].
Introducing canopy cover (Cover10) into the model did not improve goodness of fit (model 3 in Table 1), because Cover10 and TCH were closely correlated (r = 0.82) and including both led to redundancy.
