Trends in International Mathematics and Science Study 2003 Nonresponse Bias Analysis

This technical report explores the extent of potential bias introduced into the U.S. TIMSS study through nonresponse on the part of schools. Data from the third cycle of TIMSS, conducted in April-June, 2003, are the basis for the analyses. The analyses compare selected characteristics likely to reflect bias in participation from participating and non-participating schools. Two forms of analysis were undertaken: a test of the independence of each school characteristic and participation status, and logistic regression in which the conditional independence of selected school characteristics as predictors of participation was examined. The investigation into nonresponse bias at the school level for U.S. TIMSS 2003 samples for grades 4 and 8 shows that there was no statistically significant relationship detected between participation status and the majority of school characteristics that are available for analysis.