Using Nonparametric Models to Forecast the Number Injuries of COVID -19

Abstract

The vast majority of countries are experiencing economic and health crises due to the current pandemic of the Coronavirus disease (COVID-19). In order to study any phenomenon, it is necessary to model the variables that we believe are influential in this phenomenon, and at the forefront of these models are the so-called the Regression Models. These models explore the relationship between the explanatory variables and the response variable. As the parametric methods assume that the sample comes from a specific population with a known family of distributions, the assumed parametric distribution is often not necessarily the actual distribution of the data to be solved, as the wrong assumption of the parametric distribution of the given data may lead the statistical methods used to incorrect conclusions and inconsistent estimates. Misusing of the parametric distribution for the given data may lead to incorrect conclusions and inconsistent estimation. Parametric methods are often inappropriate for data that is small or has no known distribution, while the nonparametric methods, which are a wide array of flexible models, can be less stringent and less descriptive, that is, they give general description for the relationship, which made it desirable tool for researchers. The objective of the study is to analyze the number of cases of (COVID-19) in Iraq using nonparametric models, such as Polynomial Regression, Spline Regression, and Generalized Additive Models GAM. These models will be compared using the comparison criterion Akaike information criterion (AIC) and Bayesian information criterion (BIC) for choosing the best model to forecast the number of COVID-19 infections in Iraq.