Estimating the bias in meta analysis estimates based on fixed effect model for data with missing variability measures

Nik Idris, Nik Ruzni (2012) Estimating the bias in meta analysis estimates based on fixed effect model for data with missing variability measures. In: IIUM Research, Invention and Innovation Exhibition, IRIIE 2012, 21 - 22 February, 2012., Cultural Activity Centre (CAC) and KAED Gallery, IIUM.. (Unpublished)

Preview	PDF (poster) - Accepted Version Download (401kB) | Preview
	PDF (certificate of participation) Restricted to Repository staff only Download (41kB) | Request a copy

Abstract

A common drawback with meta analysis is when the variability measures, particularly the variances , are not reported, or “missing” in the individual study. Among the approaches adopted in handling this problem is through exclusion of the studies with missing variances. Alternatively, the missing study-variances could be imputed. This paper examines the analytical implications of these two approaches on the overall effect estimate and the corresponding variances. The bias in these estimates are derived using the Fixed Effect model. The results show that no bias is expected in the estimate of the overall effect using both approaches. Similarly, there is no bias in the variance of the effect estimate when the missing study-variances are imputed and homogeneous study-variances are assumed across the studies. However, if the magnitude of the missing study-variances are mostly larger than those that are reported, imputation leads to under estimation of the variance of the effect estimate. This is a likely case in meta analysis. When studies with missing variances were excluded from analysis, the variances of the effect estimate are overestimated, and the magnitude of the bias in this case is relatively larger when compared to those from complete imputed data.

Item Type:	Conference or Workshop Item (Poster)
Additional Information:	4059/25596
Subjects:	Q Science > Q Science (General)
Kulliyyahs/Centres/Divisions/Institutes (Can select more than one option. Press CONTROL button):	Kulliyyah of Science > Department of Computational and Theoretical Sciences
Depositing User:	Dr Nik Ruzni Nik Idris
Date Deposited:	17 Dec 2012 16:02
Last Modified:	17 Dec 2012 16:02
URI:	http://irep.iium.edu.my/id/eprint/25596

Actions (login required)

View Item