Methods For The Analysis Of Explanatory Linear Regression Models With Missing Data At Random

Abstract

The analysis of missing data is done with techniques design for the data set that are completely missing at random as complete case analysis, regression imputation, multiple imputation and maximum likelihood based method .. however , the data condition required to minimize the bias derived from an incorrect analysis have not been full determined. In present work , a simulated data have been carried out to establish the best strategy of analysis for multiple regression applicable to the data set with values missing at random. The results shows that the smallest bias is obtained with maximum likelihood based methods and multiple imputation techniques, although the low percentage of missing data, and sample size have no effect on the bias of single imputation and complete case analysis.