Do instructional attributes pose multicollinearity problems? An empirical exploration

Alauddin, Mohammad and Nghiem, Hong Son (2010) Do instructional attributes pose multicollinearity problems? An empirical exploration. Economic Analysis and Policy, 40 3: 351-361.

Attached Files (Some files may be inaccessible until you login with your UQ eSpace credentials)
Name Description MIMEType Size Downloads
UQ223922_OA.pdf Full text (open access) application/pdf 162.15KB 0
Author Alauddin, Mohammad
Nghiem, Hong Son
Title Do instructional attributes pose multicollinearity problems? An empirical exploration
Journal name Economic Analysis and Policy   Check publisher's open access policy
ISSN 0313-5926
Publication date 2010-12-11
Year available 2010
Sub-type Article (original research)
Open Access Status File (Publisher version)
Volume 40
Issue 3
Start page 351
End page 361
Total pages 11
Place of publication Brisbane, Qld, Australia
Publisher Economic Society of Australia and New Zealand. Queensland Branch
Collection year 2011
Language eng
Abstract It is commonly perceived that variables ‘measuring’ different dimensions of teaching (construed as instructional attributes) used in student evaluation of teaching (SET) questionnaires are so highly correlated that they pose a serious multicollinearity problem for quantitative analysis including regression analysis. Using nearly 12000 individual student responses to SET questionnaires and ten key dimensions of teaching and 25 courses at various undergraduate and postgraduate levels for multiple years at a large Australian university, this paper investigates whether this is indeed the case and if so under what circumstances. This paper tests this proposition first by examining variance inflation factors (VIFs), across courses, levels and over time using individual responses; and secondly by using class averages. In the first instance, the paper finds no sustainable evidence of multicollinearity. While, there were one or two isolated cases of VIFs marginally exceeding the conservative threshold of 5, in no cases did the VIFs for any of the instructional attributes come anywhere close to the high threshold value of 10. In the second instance, however, the paper finds that the attributes are highly correlated as all the VIFs exceed 10. These findings have two implications: (a) given the ordinal nature of the data ordered probit analysis using individual student responses can be employed to quantify the impact of instructional attributes on TEVAL score; (b) Data based on class averages cannot be used for probit analysis. An illustrative exercise using level 2 undergraduate courses data suggests higher TEVAL scores depend first and foremost on improving explanation, presentation, and organization of lecture materials.
Q-Index Code C1
Q-Index Status Confirmed Code
Institutional Status UQ

Document type: Journal Article
Sub-type: Article (original research)
Collections: Official 2011 Collection
School of Economics Publications
School of Medicine Publications
Version Filter Type
Citation counts: Google Scholar Search Google Scholar
Created: Wed, 08 Dec 2010, 10:50:02 EST by Alys Hohnen on behalf of School of Economics