Minggu, 05 Mei 2013

Comparison of Regression Methods end Expectation Maximization (EM) Fill In Missing data ( Study data anthropometry Wisma Indah Toddeler Health Center District Bojonegoro )


The missing data is the problem which happen in reseaerch that is caused by some factors. In large amount, missing data can influence the validity of research anlysis result. Missing value analysis with regression and EM method is one of methods to estimate missing data. The purpose of this study was to compare the regression and EM methods in estimating missing data values.

This type of research was non-reactive with secondary data analysis. The variables analyzed were age, height and weight of infants in health centers Wisma Indah of Bojonegoro regency. Data that was taken consist of 500 infants. The first prosedur was lossing data with simulation data at 20% then performed with data imputation with the EM and regression methods to replicate as much as three times. To find the difference of the original data with the results of estimation was tested with the the same subject anova. The best method was determined by looking at the closeness of the highest correlation and the average square of the smallest difference.

Results showed both regression and EM methods no significant differences in mean values and standard deviations. the regression method, a good method was regression with non Adjustment with 2 predictors, the EM method, a good method was EM with 2 predictors and 66.66% for EM methods had on average than the least squares regression methods vary, so it could be interpreted EM method better than the regression method in estimating the missing data.

EM method used maximum likelihood approach with iteration process until the value going convergen.

Key word : Regression, EM, Missing data
Related Posts Plugin for WordPress, Blogger...

Dalam Lingkaran Kami