Using Different Threshold Value in Comparison Some of Methods Wavelet Estimation for Non Parametric Regression Function with Missing Data

Abstract

The problem of missing of some of sample observations is one of the main problems that face researcher during the statistical analysis , the main problem of missing data are as follows damage , negligence, death and morbidity as in the case of clinical studies The presence of such a problem within the data may influence on the analysis and accordingly it may lead to misleading conclusions despite the fact that the wavelet estimations are of high efficiency in estimating the regression function , but it may be influenced by the problem of missing data , in addition to the impact of the problem of miss of accuracy estimation it is not possible to apply these methods because of the miss of one of its conditions which is dyadic sample size .According to the great impact stem from that problem , many researchers who devoted their study to process this problem by using traditional methods in processing missing data , where as the researcher used imputation methods more efficient and effective to process the missing data as a primary stage so that these data will be ready and available to wavelet application, as a result simulation experiment proved that the suggested methods (DRPW) are more efficient and superior to other methods , this paper also includes the auto correction of boundaries problem by using polynomial models , and using different threshold values in wavelet estimations , SINCE the suitable choice of this value is decisive accuracy of these estimations