A Comparison of Two Methods for Analyzing Data from a Methodological Study of Paper-and-Pencil vs. Web-Based Survey Administration: Sample Survey and Group Randomized Trial

• Activity Number: 455
• Type: Topic Contributed
• Date/Time: Wednesday, August 4, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Maxine M. Denniston*+ and Danice K. Eaton and Sherri L. Pals
• Companies: CDC and CDC and CDC
• Keywords: group randomized trial ; sample survey ; data analysis ; web-based ; paper-and-pencil

Abstract

In 2008 CDC conducted a study comparing web-based vs. paper-and-pencil administration of the Youth Risk Behavior Survey, a biennial national survey of U.S. high school students that assesses priority health-risk behaviors. Although analysis as a sample survey is appropriate because students were clustered within classrooms, analysis as a group randomized trial (GRT) also is appropriate because study condition was assigned to intact classrooms. To decide which analytic approach to use, we compared results for risk behavior prevalence from SUDAAN with those from a GRT analysis using SAS PROC GENMOD. Adjusted Odds Ratios from the 2 analyses were similar and significance at the .05 level differed for only 2 of 74 risk behaviors assessed. We conclude that for at least some studies which could be analyzed using sample survey or GRT, the results will be equivalent regardless of the method used.