comparison Bennett's inequality and regression in determining the optimum sample size for estimating the Net Reclassification Index (NRI) using simulation

Abstract

Researchers have increased interest in recent years in determining the optimum sample size to obtain sufficient accuracy and estimation and to obtain high-precision parameters in order to evaluate a large number of tests in the field of diagnosis at the same time. In this research, two methods were used to determine the optimum sample size to estimate the parameters of high-dimensional data. These methods are the Bennett inequality method and the regression method. The nonlinear logistic regression model is estimated by the size of each sampling method in high-dimensional data using artificial intelligence, which is the method of artificial neural network (ANN) as it gives a high-precision estimate commensurate with the data type and type of medical study. The probabilistic values obtained from the artificial neural network are used to calculate the net reclassification index (NRI). A program was written for this purpose using the statistical programming language (R), where the mean maximum absolute error criterion (MME) of the net reclassification network index (NRI) was used to compare the methods of specifying the sample size and the presence of the number of different default parameters in light of the value of a specific error margin (ε). To verify the performance of the methods using the comparison criteria above were the most important conclusions were that the Bennett inequality method is the best in determining the optimum sample size according to the number of default parameters and the error margin value