We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
A Three-stage Estimator for Studies with Repeated and Possibly Missing Binary Outcomes.
- Authors
Lipsitz, Stuart R.; Laird, Nan M.; Harrington, David P.
- Abstract
This paper describes methods for analysing repeated measures data when the outcome is binary at each occasion. The models relate the expected value of the binary response at each occasion to covariates via the logistic link function. The method of estimation is a three-stage, generalized least squares method and is similar to the methods described by Ware and Jennrich and Schluchter for repeated continuous outcomes. The estimates are asymptotically equivalent to the estimates obtained from Liang and Zeger's 'generalized estimating equations' but have the advantage that they exist in some situations (i.e. small sample sizes, unbalanced data) when the Liang and Zeger estimates do not and can be computed using a newly available statistical package (BMDP 5V) designed for analysing unbalanced repeated continuous responses. Residual diagnostics are also briefly discussed.
- Subjects
ESTIMATION theory; LEAST squares; BINARY number system
- Publication
Journal of the Royal Statistical Society: Series C (Applied Statistics), 1992, Vol 41, Issue 1, p203
- ISSN
0035-9254
- Publication type
Article
- DOI
10.2307/2347629