Reduced and simple models were weaker, explaining 48 79% and 42 79%, respectively, of the variance among descriptors.
Although AIC favored model HET, support for the HET model was weaker than for the HOM model when a more severe penalty for complexity (BIC) was applied.
However, to date, the performance of fall risk models is weak compared with models estimating, for example, cardiovascular risk.
However, according to AIC all other models are weak (zero values for Akaike weights) and should not be considered, a result that could not be attained using R 2 values alone.
Here, as is often customary, the introduced models are weak-sense stationary (WSS) random processes [ 13– 15].
Each of these lasso models is weak in model selection as it suffers from the model selection inconsistency of the lasso.
They also concurred that the data on mouse models are weak, primarily due to their marginal/questionable significance (see detailed comments below).
Multicollinearity analysis showed that, with the exception of the correlation between admission diagnostic categories and length of ICU stay (0.47), all other correlations between the independent variables and coefficient estimates included in the regression models were weak (below 0.30).
In all regression models except for "Energy" and "Emotional Reaction", Durbin-Watson statistics were in an acceptable range between 1.6 ("Social Isolation") and 2.09 ("Sleep"), However the explanatory power of the regression models is weak: Only between 2% and 7% of the variation in the domains of HRQOL can be explained by age, gender and CDS.
Multicollinearity analysis showed that, with the exception of the correlation between admission diagnostic categories and length of ICU stay (0.47), all other correlations between the independent variables and coefficient estimates included in the five regression models were weak (below 0.30; data not shown).
