"Missing Data: Analysis and Design in Psychological Trauma Research"
The presentation will be in two parts.
In Part 1, I will discuss theory and practice in missing data analysis, and
will provide a brief demonstration of two acceptable approaches to analysis
with missing data: multiple imputation and full-information maximum likelihood
(FIML) with structural equation modeling. In Part 2, I will present two kinds
of planned missing data designs that may be useful in trauma research. One design,
called the 3-form design, is a member of a family of designs that can improve
the efficiency of one's survey measurement. By building in some missingness
by design, the researcher can actually test more hypotheses with reasonable
power. Scenarios relevant to trauma research are presented to illustrate the
benefits of the design. The second design involves the use of two methods of
measurement. One method is an inexpensive, but less valid, measure of the construct
of interest, and is typically administered to a large number of respondents;
the second method is an expensive and more valid measure of the construct that
is administered only to a subset of the respondents. The potential benefits
of the procedure are illustrated with a trauma-relevant example. The presention
will conclude with a question and answer session.