Prediction of the NASH through penalized mixture of logistic regression models

In this paper an appropriate and interpretable diagnosis statistical model is proposed to predict Nonalcoholic Steatohepatitis (NASH) from near infrared spectrometry data. disease, unknown patients’ profiles are expected lead a different diagnosis. The has then take into account the heterogeneity of data dimension spectrometric To end, we propose fit mixture on joint distribution diagnostic binary variable covariates selected in spectra. penalized maximum likelihood estimator considered. practice, twofold penalty both regression coefficients covariance parameters imposed. Automatic selection criteria, such as AIC BIC, used select amount shrinkage number clusters. performance overall procedure evaluated by simulation study, its application NASH set analyzed. leads better prediction than competitive methods provides highly results.