ESTIMATED NON-PARAMETRIC AND SEMI-PARAMETRIC MODEL FOR LONGITUDINAL DATA

This type of data (longitudinal data) has a large informational content for the phenomenon studied and thus enables us to obtain accurate estimates, and longitudinal data can be defined as data that combines between time series data and cross-sectional data, meaning that it studies cross-sectional data and its movements during a certain period of time and from here the importance of the research is highlighted in the comparison between the preference of the longitudinal data models on the one hand, as well as the preference for the nonparametric and semi-parametric models through which it is possible to describe the nature of the relationship between the model variables. The research aims to estimate a model for parametric and semi-parametric that accurately describes the nature of the relationship between the value of industrial production for ten large industrial establishments in the public sector in Iraq for the period of time (2010-2018) and both the value of the requirements and numbers of workers in those establishments, and to achieve that aims, a method was employed Local Linear Polynomial regression for estimating the nonparametric model and Profile Least Square method for estimating the semi-parametric model, and the two methods included employing Local Linear Polynomial (regression with different degrees and the (Gaussian) and (Epanchnikov) kernel functions. As well as making a comparison between the two models through the use of three criteria for comparison, which are the mean squares of error (MSE), the mean of absolute deviations standard (MDAE), and the mean of absolute deviations (MAE). The results obtained showed that the estimation method Profile Least Square is better than Local Linear Polynomial Estimation.