Advances and Applications in Statistics
Volume 17, Issue 2, Pages 175 - 190
(August 2010)
|
|
HETEROGENEITY OF ODDS RATIOS OVER TIME IN CROSS-SECTIONAL TREND STUDIES: AN OVERWEIGHT-OBESITY EXAMPLE FROM A NATIONWIDE POPULATION-BASED STUDY
Wei Yang, Fares Qeadan, Daniel Cook and Julie Smith-Gagen
|
Abstract: To ascertain whether measures from studies are consistent, statistical tests of heterogeneity are commonly used through meta-analyses. The two foremost used heterogeneity tests are the Q Cochran’s heterogeneity statistic and I2 measure. However, for population-based cross-sectional trend studies, the compatibility of the measurements over time is ensured and therefore conducting meta-analysis over time would provide several benefits. Series of cross-sectional surveys are appealing in research to track long-term changes in health behaviors. The most materialized example, in health data, of such studies is the Behavioral Risk Factor Surveillance System (BRFSS). The design and methods for the BRFSS over time are almost identical with equivalent sampling and measurement approaches that exceedingly tone with the utilization of Cochran’s heterogeneity statistic and I2. We examine the heterogeneity of odds ratio over time for the risk factors of overweight/obesity in the United States for the odd years since 1991-2007. Data was obtained from the nationwide BRFSS and processed via multiple logistic models for each year. Odds ratios for the risk factors, their 95% confidence interval over time, Cochran’s heterogeneity statistic and I2 measure were computed. Moreover, when heterogeneity of odds ratio was identified, a follow up periodization analysis was conducted. The results of this study indicate that odds ratios of smoking status, sex, diabetes, race, marital status and physical activity for overweight/obesity are heterogeneous over time, and their time-inconsistency diminishes before and after the year 1999. |
Keywords and phrases: heterogeneity, meta-analysis, BRFSS, trend studies, logistic. |
|
Number of Downloads: 343 | Number of Views: 1197 |
|