Application of semi-parametric single-index two-part regression and parametric two-part regression in estimation of...

For the purpose of cost modeling, the semi-parametric single-index two-part model was utilized in the paper. Furthermore, as functional gastrointestinal diseases which are well-known as common causes of illness among the society people in terms of both the number of patients and prevalence in a specific time interval, this research estimated the average cost of functional gastrointestinal diseases.

Health care policy-makers seek for real and accurate estimations of society's future medical costs. However, data dealt with in hygienic studies have characteristics which make their analysis complicated; distribution of cost data is highly skewed since many patients pay great costs. In addition, medical costs of many persons are zero in a specific time interval. Indeed, medical costs data are often right skewed, including remarkable number of zeros, and may be distributed non-homogeneously.

In modeling these costs by the semi-parametric single-index two-part model, parameters were determined by method of least squares; a result of this method was compared with the results yielded from two-part parametric model.

Average costs of functional gastrointestinal diseases and their standard deviation in semi-parametric and parametric methods were yielded as $72.69±108.96 (R(2)=0.38) and $75.93±122.29 (R(2)=0.33) respectively.

Based on R(2) index, the semi-parametric model is recognized as the best model. Totally, the two-part parametric regression model is a simple and available model which can be easily interpreted; on the other hand, though the single-index two-part semi-parametric model cannot be easily interpreted, it has considerable flexibility. The study goals can be indeed used as the main factor for choosing one of these two models.