To test whether these three groups are outstanding, label shuffling of the groups and retrying classification with the four best variables was performed.
Therefore, group labels were shuffled randomly 500 times, each time random forest was calculated, the variables were sorted by relevance and the 4 best variables were then used for recalculation of random forest with same shuffled groups.
The best variables were then plotted on the same ROC curve and the area under the curve (AUC) values were directly compared to assess whether one measure was superior to the other for differentiating the groups.
The best variable is defined as the variable within a factor that was most strongly associated with the adjusted injury rate, using manual backwards stepwise regression.
The model was run with forward entry where the best variable is entered, then the next, etc., until no more variables significantly contribute to the prediction.
When a unique or best variable is identified, it could be renamed straightaway into the dataset at runtime; however, instead, it is recorded in '_Structure.dta' for keeping track and documenting the changes operated in the datasets ('_Structure.dta' can be used to describe the data used in statistical analyses).
PCA works best when variables are highly correlated and their distribution varies across households [ 13].
The final set of specific evidenced based best practice variables are shown in Table 2.> The primary outcome measure for the study was a composite score that consisted of all the best practice variables from the PHR and was measured out of 31 (in Phase 1) and in the EHR was out of 28 (in Phase 2).
Firstly, similarities can be drawn between this approach and the wrapper approach to feature selection [ 13], where the best features, or variables, are chosen for classification based on how well they work accumulatively with a given classifier.
The best predictor variables were selected (7 topographic and 5 climatic variables) using the Least Absolute Shrinkage and Selection Operator method.
