Apologies for cross-posting.
Geoff Der <[log in to unmask]> found the reference I mentioned in my
last posting, namely:
DOCUMENT TYPE
Journal-Article
TITLE
Maximizing the usefulness of data obtained with planned missing
value patterns: An application of maximum likelihood procedures.
AUTHOR
Graham,-John-W.; Hofer,-Scott-M.; MacKinnon,-David-P.
SOURCE
Multivariate-Behavioral-Research.1996; Vol 31(2): 197-218.
ISSN
0027-3171
PUBLICATION YEAR
1996
ABSTRACT
Uses a simulation example to show that maximum likelihood (ML)
approaches and associated multiple imputations of missing data
provide unbiased, efficient estimates of variance and covariance
parameters in normally distributed and skewed data. The authors
also examine parameter estimates obtained from mean replacement,
single imputation, and pairwise deletion when data are missing
completely at random (MCAR) and where nonrandom missingness is
overlaid on the 3-form design. An example from an alcohol and drug
abuse prevention program is analyzed using ML and pairwise
estimation. A simulation is performed to indicate when it is
better to choose to split scales across forms or keep scales
complete within forms. Results show that ML estimation and
multiple imputation methods produce the most efficient and least
biased variance and covariance estimates for normally distributed
and slightly skewed data when data are MCAR. ((c) 1997
APA/PsycINFO, all rights reserved) .
Thanks, Geoff. Mark.
---------------------------------------------------------------
Dr Mark S Gilthorpe, Senior Lecturer
National Centre for Transcultural Oral Health
Eastman Dental Institute for Oral Health Care Science
University College London, 256 Gray's Inn Road, London WC1X 8LD
Tel: ++44(0)171 915 1234 Dept: ++44(0)171 915 1193
Fax: ++44(0)171 915 1233 mailto:[log in to unmask]
Eastman: http://www.eastman.ucl.ac.uk
Dept: http://www.eastman.ucl.ac.uk/depart/toh/index.html
---------------------------------------------------------------
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|