Annual Meeting Recordings are now available for purchase
302796
Handling missing data for PLS path model in health research: A comparison study
Tuesday, November 18, 2014
Shengtian Hou
,
School of Management, Beijing University of Chinese Medicine, Beijing, China
Yongkang Zhang
,
Department of Global Health Systems and Development, Tulane University, New Orleans, LA
Partial least square (PLS) path modeling has been widely used in health research area owing to its effectiveness in analyzing the cause-effect relations between latent variables. However, few discussions can be found on how to handle missing data when using PLS path model. Missing data are a recurring problem in health-related surveys. Improper handling for missing value can lead to bias results. Several software, like SmartPls, PLS-GUI and PLS-Graph have been developed specifically for PLS path modeling but they just offer limited missing data handling methods like listwise deletion, mean imputation which are supposed to be invalid. In this study, a PLS path model is constructed to analyze the influencing factors of after-sale service satisfaction for medical equipment. The data were collected from 739 hospitals in China. The proportion of missingness of 25 measurement indicators is from 0.3% to 74.9%. We adopted five different missing data handling methods, including listwise deletion, mean imputation, regression imputation (RI), expectation maximization (EM) and multiple imputations (MI) to treat missing value. The results suggested that there are no differences in the validity and reliability of the PLS model when using different missing data handling methods. Better goodness-of-fit was achieved when using RI, EM and MI method. The same results, which are consistent with results of qualitative study, were gotten when it comes to the significant influencing factors of after-sale service satisfaction by using RI, EM and MI. In conclusion, RI, EM and MI are appropriate to process missing data when using PLS path model.
Learning Areas:
Administration, management, leadership
Biostatistics, economics
Conduct evaluation related to programs, research, and other areas of practice
Learning Objectives:
Compare different methods for handling missing data when using PLS path model to analyze cause-effect relations in health marketing research. Identify the differences of results by using listwise deletion, mean imputation, regression imputation, expectation maximization and multiple imputations. Clarify whether the results are consistent with qualitative study results and other information in order to evaluate the effectiveness of different missing data handling methods.
Keyword(s): Methodology, Marketing
Presenting author's disclosure statement:Qualified on the content I am responsible for because: I have been the co-principal of funded grants focusing on patient satisfaction and corporate social responsibility. I also was the co-author for several peer-reviewed journal papers.
Any relevant financial relationships? No
I agree to comply with the American Public Health Association Conflict of Interest and Commercial Support Guidelines,
and to disclose to the participants any off-label or experimental uses of a commercial product or service discussed
in my presentation.
