A Mixture Regression Approach for Modelling Early Post Operative Hypocalcemia

Regression Approach for Modelling Early Post Operative Hypocalcemia

In this paper, we introduce a new generalization of univariate and bivariate modified discrete Weibull distribution. Various properties of the univariate generalized modified discrete Weibull distribution, such as survival function, probability mass function, hazard rate function, probability generating function, and moment generating function, are derived. The joint distribution function, joint probability mass function, marginal distributions, moment generating function, and conditional distribution of the proposed bivariate distribution are derived. Parameters of the distributions are estimated using maximum likelihood estimation. The use of these distributions is illustrated using real-life datasets.

In this paper, we show that a two component Laplace mixture model is an appropriate distribution to model postoperative calcium levels of patients undergoing thyroidectomy or surgery for thyroid diseases. A mixture regression model is constructed to predict postoperative calcium levels based on a pre-determined set of potential predictors. The parameters of the model are estimated using EM algorithm. Our study based on a real data set shows that two component Laplace mixture regression model is suitable for prediction and interpretation compared to the usual Gaussian mixture regression model.