Evaluating the Performances of Missing Data Handling Methods in Ability Estimation From Sparse Data.
Large amounts of missing data could distort item parameter estimation and lead to biased ability estimates in educational assessments. Therefore, missing responses should be handled properly before estimating any parameters. In this study, two Monte Carlo simulation studies were conducted to compare...
| Published in: | Educational & Psychological Measurement Vol. 80; no. 5; pp. 932 - 955 |
|---|---|
| Main Authors: | , |
| Format: | Article |
| Published: |
Sage Publications Inc.
Oct2020
|
| Subjects: | |
| Online Access: | View this record in EBSCOhost |
| fields | @attributes: recordID: 1 pdfLink: plink: https://search.ebscohost.com/login.aspx?direct=true&db=ssf&AN=145141367&site=ehost-live header: @attributes: shortDbName: ssf uiTerm: 145141367 longDbName: Social Sciences Full Text (H.W. Wilson) uiTag: AN controlInfo: bkinfo: jinfo: jid: 00131644 EPM jtl: Educational & Psychological Measurement issn: 00131644 maglogo: Y pubinfo: dt: Oct2020 vid: 80 iid: 5 pid: 344 pub: Sage Publications Inc. artinfo: ui: 145141367 10.1177/0013164420911136 ppf: 932 ppct: 23 formats: tig: atl: Evaluating the Performances of Missing Data Handling Methods in Ability Estimation From Sparse Data. aug: au: Xiao, Jiaying Bulut, Okan affil: University of Alberta, Edmonton, Alberta, Canada su: Computer simulation Database management Database evaluation Statistical sampling Statistics Data analysis Research bias Random forest algorithms sug: subj: Computer simulation Data Processing, Hosting, and Related Services Marketing Research and Public Opinion Polling Database management Database evaluation Statistical sampling Statistics Data analysis Research bias Random forest algorithms keyword: ability estimation CART missing data multiple imputation random forest ability estimation CART missing data multiple imputation random forest ab: Large amounts of missing data could distort item parameter estimation and lead to biased ability estimates in educational assessments. Therefore, missing responses should be handled properly before estimating any parameters. In this study, two Monte Carlo simulation studies were conducted to compare the performance of four methods in handling missing data when estimating ability parameters. The methods were full-information maximum likelihood (FIML), zero replacement, and multiple imputation with chain equations utilizing classification and regression trees (MICE-CART) and random forest imputation (MICE-RFI). For the two imputation methods, missing responses were considered as a valid response category to enhance the accuracy of imputations. Bias, root mean square error, and the correlation between true ability parameters and estimated ability parameters were used to evaluate the accuracy of ability estimates for each method. Results indicated that FIML outperformed the other methods under most conditions. Zero replacement yielded accurate ability estimates when missing proportions were very high. The performances of MICE-CART and MICE-RFI were quite similar but these two methods appeared to be affected differently by the missing data mechanism. As the number of items increased and missing proportions decreased, all the methods performed better. In addition, the information on missing data could improve the performance of MICE-RFI and MICE-CART when the data set is sparse and the missing data mechanism is missing at random. pubtype: Academic Journal doctype: Article src: R language: English refInfo: copyright: @attributes: flag: N holdings: @attributes: islocal: N |
|---|