Practical application of meta-analysis results: avoiding the double use of data

Abstract

Meta-analysis is an important new tool for synthesizing scientific knowledge from many previous studies. In fisheries, meta-analyses can be used to obtain prior distributions or penalty functions for parameters used in stock assessment models. Two types of results are generally published in a meta-analysis: Type A, the updated results for each stock used in the meta-analysis, and Type B, the results that would best describe a new stock. Including these results in assessments for the individual stocks would result in double use of the data if the assessments include the input data used in the meta-analyses, which they typically would. To solve this problem, we recommend that an additional form of results should be reported in meta-analyses: Type C, the results for a new stock obtained by sequentially excluding each stock's data set and repeating the meta-analysis. Type C results should be used whenever the assessment input data overlap with the meta-analysis input data, avoiding the double use of data. We illustrate the impact of this reporting change on the results of a recent meta-analysis.