Testing the Task and Ego Orientation in Sport Questionnaire (TESOSQ) Measurement Model With Incomplete Data: An Application of Maximum Likelihood-Based Estimation Procedures
Titel:
Testing the Task and Ego Orientation in Sport Questionnaire (TESOSQ) Measurement Model With Incomplete Data: An Application of Maximum Likelihood-Based Estimation Procedures
Auteur:
Li, Fuzhong Duncan, Terry E. Duncan, Susan C. Harmer, Peter Acock, Alan
Verschenen in:
Measurement in physical education & exercise science
Paginering:
Jaargang 2 (1998) nr. 1 pagina's 1-19
Jaar:
1998-03-01
Inhoud:
Nearly all psychosocial survey datasets contain missing data. Recent advances in statistical theory, and implementation of the theory in modeling software, have made it possible to handle various types of missing data with the proviso that data are missing at random. This study presents an application of a confirmatory factor analysis with incomplete data, demonstrating the utility of several maximum likelihood-based missing data estimation procedures: multigroup, single group, and expectation maximization multiple-imputation. The substantive example used to demonstrate these procedures was a 2-factor measurement model of achievement goal orientation in sports with a sample size of 1,002. Of the total sample, 124 participants (12%) had incomplete data in the form of nonresponse. Initial analysis showed that the data missing were completely random. Subsequent modeling of the incomplete data revealed that, across the 3 data estimation procedures, parameter estimates were remarkably similar and standard errors of estimates were small in magnitude, suggesting the consistency of these procedures in analyzing incomplete data.