The models using the ordinal outcomes did not lead to appreciably different conclusions than models using binary outcomes.
The independence of the predictors of the MetS was tested in the multivariate models using binary logistic regression.
The correlation between metabolic syndrome and serum leptin quartiles was calculated independent of age, tobacco usage, and BMI in three models using binary logistic regression analysis (Table 4).
eAUC are 0.79 and 0.80 for age-adjusted model and fully-adjusted model respectively; Fully-adjusted models used binary logistic regression, adjusted for age, BMI, smoking, drinking, and education.
bAreas under the receiver operating characteristic curves (AUC) are 0.61 for age-adjusted model and 0.69 for fully-adjusted model respectively; Fully-adjusted models used binary logistic regression, adjusted for age, BMI, smoking, drinking, education, and baseline falls.
Additionally, we also developed SVM based models using binary profile where each position is represented by a vector of dimension of 20 (each element represent presence or absence of a specific type of residue).
Final models using binary logistic regression for analyzing betel-nut and cigarette consumption were chosen, based on likely and relevant confounders after univariate analysis (candidate correlates were excluded from the final analyses if p > 0.05 in the univariate analysis with outcome variable).
The proposed model uses binary strings which represent the state of the network switches and capacitors.
This particular situation cannot be modeled using binary interaction parameters only, in contrast to the results obtained with acetone as the precipitant.
We further developed SVM model using binary profiles of patterns as input features.
