Both of the AUCs of H-FABP+PSI and H-FABP+CURB-65 were higher than that of H-FABP, PSI, or CURB-65 alone, but the differences of AUCs between the predictors were not significant.
Except for the dosage of noradrenaline the mortality predictors were not different between the three groups.
VIF's which are less than 5 indicate that covariation between the predictors is not strong enough to violate the assumptions of the model (Bowerman and O'Connell 1990; Myers 1990).
However, a direct head-to-head comparison between both predictors was not performed despite that fact that the concordance (i.e., κ score) between these two predictors was 0.59 (0.58–0.61), which is considered moderate agreement and similar to the κ scores obtained when histological grade is evaluated by two independent observers [ 7].
Although our predictor achieves the best performance, the difference between our predictor and other predictors is not statistically significant by using pairwise Student's t-test, partly due to the small number of Liang's independent datasets [ 13, 14].
The Lasso has been shown to be consistent for model selection under certain conditions given that the correlation between relevant (true) and irrelevant predictors is not too large [Zhao and Yu, 2006], or in other words, the selected model will contain the true model with high probability.
The spatial resolution of the predictors was not homogeneous between factors and thus all the ecogeographical information was finally translated into UTM 10×10 km squares using the Extract module of the Idrisi Andes software package.
This indicates that provided one has evidence of moderate to tight regulation between the genes, and the number of predictors is not too large, one should use the CoD estimator based on resubstitution.
The authors present numerical experiments that compare the performance of several CoD estimators and conclude that the empirical (resubstitution) CoD can be the best choice, provided that one has evidence of moderate to tight regulation between the genes, and the number of predictors is not too large.
However, the linear assumption between the dependent variable and both continuous predictors is not necessarily met in all situations, especially for temperature and air pollution in a multisite study with variation in study populations.
