> List members, > > I am planning a large telephone survey. It may not be surprising that I > am interested in assessing more constructs than I may be able to fit > within a reasonable amount of time. Although reducing the number of > variables is a possibility, I vaguely recall hearing of the possibility of > using planned missingness, and then imputing for full sample analyses. I'd > have a core set of questions for everyone, and then randomly assign > individuals to alternate forms of the questionnaire that have certain > sections but not others. > > If my memory isn't completely off, can anyone recommend readings on how to > implement such a plan, how to impute the missing data for full sample > analyses, and the costs/benefits of such an approach.
Mike -- Try Graham, J. W., Hofer, S. M., & MacKinnon, D. P. (1996). Maximizing the usefulness of data obtained with planned missing value patterns: An application of maximum likelihood procedures. _Multivariate Behavioral Research, 31_ 197-218. It seems to fit the bill. > Thanks in advance. > > Mike Frone HTH, Pat Malone -- Patrick S. Malone, Ph.D., Research Associate Duke University Center for Child and Family Policy Durham, North Carolina, USA e-mail: [email protected] http://www.duke.edu/~malone/
