Division: Workshops

Title: Meta-Analysis for the Synthesis of Evidence in Agriculture

Lead Community Sponsor:
Cosponsor: Biometry and Statistical Computing
Community Cosponsor: Statistical Education/Training for Researchers Community
Format:
Keywords:

Session Description: Meta-analysis is the analysis of the results of multiple independent studies, which is typically performed in order to synthesize the evidence from many possible sources in a formal quantitative manner. In the simplest and most common case, the outcome of each study becomes a single observation in the meta-analysis of all available studies. The discipline developed originally in the social sciences in the 1970s, based on earlier pioneering contributions by Fisher Pearson, Yates, and Cochran, and has now been embraced within many scientific disciplines, especially in medical research. Since 1980, over 80,000 journal articles have been published that either use the method or expand on the procedure. After a very slow start, agricultural scientists are now applying meta-analytic methods to a range of problems.  This workshop will cover basic concepts and approaches in meta-analysis, and show how to estimate parameters and interpret results. Common methods for expanding the meta-analytical model to account for study-level characteristics (so called moderator variables) and for multiple treatments per study will be reviewed. The advantages of meta-analysis will be presented in terms of prediction and high statistical power; the major (potential) disadvantage, bias, will also be discussed. Methods will be explained using SAS software, including the use of specialized macros. R code will also be shown for some applications. Attendees should bring a laptop with SAS 9.3 or later, or the free SAS University Edition, installed.