Determination of the risk factors affecting the mean arterial blood pressure of pregnants via varying coefficient regression model

Varying Coefficient Models are used to reveal properties of multidimensional data among nonparametric regression models. Since these models allow the coefficients to change over the variable called the "index variable", the curse of dimensionality is avoided. Multidimensionality prevents nonparametric models from making an effective prediction. In varying coefficient model, it can be considered that the coefficients are changed in each layer of the index variable and smoothing is done by this way. From this, it is understood that there are nonlinear relationships between index variable(s) and independent variables. Varying Coefficient Model is an alternative to additive models. As it permits the vary of model coefficients over some of the variables, such as time, temperature, geographical locations, etc. it has a large application area. In this thesis study, firstly, the nonparametric regression models and estimation methods were given, then the variable coefficient regression model was introduced and the estimation methods examined. The varying coefficient regression model was applied to the data related to 450 pregnants coming the medical examination in 2018 during the April-July. The data were obtained through patient follow-up forms. Mean arterial blood pressure (MAP) variable was considered as dependent variable and gestational age considered as index variable, the number of previous pregnancies, family history of hypertension, body mass index (BMI), age, diabetes mellitus, bad obstetric history, hematocrit (HCT) and blood platelet count (BPC) were taken as independent variables. The relationship between MAP and other independent variables was analyzed via varying coefficient model as gestational age taking index variable. Key Words: Nonparametric regression models, Varying Coefficient Model, Penalized splines, Smoothing splines, Local polinomial estimation