Evaluation of Multiple Imputation in Missing Data Analysis: An Application on Repeated Measurement Data in Animal Science
Keywords:Missing observation structures, Multiple imputation, Repeated data, Mixed model, Norduz sheep
AbstractThe purpose of this study was to evaluate the performance of multiple imputation method in case that missing observation structure is at random and completely at random from the approach of general linear mixed model. The application data of study was consisted of a total 77 heads of Norduz ram lambs at 7 months of age. After slaughtering, pH values measured at five different time points were determined as dependent variable. In addition, hot carcass weight, muscle glycogen level and fasting durations were included as independent variables in the model. In the dependent variable without missing observation, two missing observation structures including Missing Completely at Random (MCAR) and Missing at Random (MAR) were created by deleting the observations at certain rations (10% and 25%). After that, in data sets that have missing observation structure, complete data sets were obtained using MI (multiple imputation). The results obtained by applying general linear mixed model to the data sets that were completed using MI method were compared to the results regarding complete data. In the mixed model which was applied to the complete data and MI data sets, results whose covariance structures were the same and parameter estimations and standard estimations were rather close to the complete data are obtained. As a result, in this study, it was ensured that reliable information was obtained in mixed model in case of choosing MI as imputation method in missing observation structure and rates of both cases.
How to Cite
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.