An application of multiple imputation using NEPS SC1 data. A comparison of R and Stata.

Autor/inn/en	Hondralis, Irina; Himbert, Elisa
Institution	Leibniz-Institut für Bildungsverläufe
Titel	An application of multiple imputation using NEPS SC1 data. A comparison of R and Stata.
Quelle	Bamberg: Leibniz-Institut für Bildungsverläufe (2018), 23 S. PDF als Volltext kostenfreie Datei Link als defekt meldenVerfügbarkeit
Reihe	LIfBi Working paper. 78
Beigaben	Literaturangaben
Sprache	englisch
Dokumenttyp	online; Monographie
Schlagwörter	Datenerhebung; Längsschnittuntersuchung; Statistische Auswertung; Antwortkontrolle; Datenmanagement; Computerunterstütztes Verfahren; Anrechnung; Daten; NEPS (National Educational Panel Study)
Abstract	This working paper describes the application of multiple imputation in the context of survey data, where missing data are a common problem. Consequently, multiple imputation methods have become a widely used approach to handle incomplete data. Despite the pervasive use of multiple imputation, more detailed and practically oriented examples and illustrations are limited. Hence, we base our user-friendly application example on data from the National Educational Panel Study SC1 (NEPS). With this working paper, we demonstrate how to apply multiple imputation with R and Stata. We further illustrate an imputation approach on how to conduct data preparation and data management prior to conducting multiple imputation. Both software programs have their advantages: R allows to specify the predictor matrix manually, which supports specific tailoring to individual data problems but can also be quite time-consuming. Both software tools offer a broad range of imputation methods, while R presents a greater variety of imputation methods based on machine learning. However, up to this date, R is less suited for longitudinal imputation, whereas we found longitudinal data preparation and imputing for longitudinal data more straight forward and easier to implement in Stata. These pitfalls should be kept in mind, when deciding on the imputation model and selecting software. (Orig.).
Erfasst von	DIPF \| Leibniz-Institut für Bildungsforschung und Bildungsinformation, Frankfurt am Main
Update	2021/2