For complex variables, such as pain, we measured multiple candidate predictor variables comprising both observational and self report tools in an attempt to optimize measurement of these factors in this population.
Those variables which were found significant in a univariate analysis were considered as candidate variables for multiple logistic regressions.
Candidate variables for multiple regression analysis were determined based on significance of factors related to GH peak determined by univariate analysis as shown in Table 2. Fasting glucose, fasting insulin, BMI, triglycerides and sex explained 54% (R = 0·5379) of the variation in GH peak (Table 3).
Independent variables which were associated with DE (p < 0.1) were entered as candidate variables into a stepwise multiple logistic regression analysis.
When changing the dependent variable by substitution of primary outcome with the total BEWE score ≥ 1 considering all surfaces, independent variables which were entered as candidate variables into a stepwise multiple logistic regression analysis were different (Table 5).
In the first step, screening bivariate ordinal logistic regression analyses were performed to identify candidate variables for the final multiple regression model.
Whilst less phylogenetic information is required to place sequences than infer a tree, Séance's accuracy suffers if there is limited resolution in the selected variable region resulting in multiple candidate nodes for placement.
Figure 1 Multiple candidate system and its PDF notation for several variables.
Candidate variables (P < 0.05) were entered into the multiple logistic regression models analysis to analyze risk factors affecting quit smoking.
As there were many candidate variables, the optimal approach we used was a stepwise multiple linear regression approach, progressively excluding variables, variables which explained minimal variance.
