Try to Raise the Predictive Precision in Binary Logistic Regression Analysis

Many reports used Functional Independence Measure (FIM) gain (FIM at discharge minus FIM at admission) because the dependent variable in multiple straight line regression analysis. Binary logistic regression analyses may also be transported out using 1 for FIM gains much like or maybe more when compared with median value and  for FIM gains beneath the median value. The deliberate conversion of quantitative FIM gains into /1 binary facts are regarded as beneficial because this doesn’t require just as much rigor based on the type or distribution of understanding.

While multiple regression analysis envisions a vertical line relationship between independent variables and dependent variable, you’ll find really many cases where no such straight line relationship exists. Especially, there’s no straight line relationship found between FIM at admission and FIM gain. Accordingly, it’s been reported that, as opposed to relying on one predictive formula, the predictive precision of motor FIM (mFIM) gain will most likely be elevated by creating two predictive formulae by stratifying mFIM scores during admission (mFIMa) into two groups

In binary logistic regression analysis, too, stratifying mFIMa to produce two predictive formulae may raise the predictive precision of mFIM gain. In addition, since you can classify independent variables in binary logistic regression analysis, it might be easy to heighten the predictive precision of mFIM gain by categorizing mFIMa.

These studies conducted binary logistic regression analysis with mFIM gain as dependent variable among stroke patients proven to convalescent rehabilitation wards in Japan. The goal of these studies ended up being compare the predictive precision of mFIM gain (a /1 binary value) between “mFIMa utilized as quantitative data”, “categorized mFIMa into 4 groups, and “progression of two predictive formulae”.

Leave a Reply

Your email address will not be published. Required fields are marked *