Performance of a Simple 3-Form Planned Missing Design in Estimating Bivariate Relations: A Monte Carlo Simulation

• In Event: AERA Poster Session 7
  In Poster Session: Division D Section 2 Poster Session 2

Abstract

Planned missing (PM) designs have been widely applied in research due to cost effectiveness and researcher-controlled missingness. This study examined a simple 3-form PM design regarding its power for bivariate correlation and test burden imposed on individual respondents, as compared to the 3-form design by Graham et al. (1996; G3F) and the full-sample (FS) design without PM data. Monte Carlo simulations of the three designs in twelve scenarios by correlations and sample sizes showed that the S3F design performed similarly to the G3F design in all scenarios tested. Both PM designs performed as well as the FS design when r ≥ 250. Simulations also provided the lowest correlations a S3F design has adequate power to estimate given four sample sizes.

Author

• Ting Dai, University of Illinois at Chicago