Application of meta-analysis in agricultural studies
Taking a final decision about the validity of a hypothesis should not be based on the results of a single experiment, mainly because results frequently vary across studies. Instead, efforts have been made to develop statistical procedures to synthesize data from one study to another. Meta-analysis is a statistical tool that allows combining the evidence from a collection of available studies on a topic of interest or specific question. It takes the outcome of each individual study, or effect size, as a single observation for a new analysis. This tool is built on the principle that individual studies, surveys, and observations contribute to the overall total knowledge base (1). There are different methods that can be used for synthesizing statistical evidence. Among the most commonly used are the method of moments, maximum likelihood (ML), restricted maximum likelihood (REML) and Bayesian. The main goal of this work is to apply these methods and compare them in a real problem to summarize and combine fertilizer recommendations for Solanaceae crops in Puerto Rico (mainly green pepper, eggplant, and tomato) (2). The advantages and disadvantages of each method are discussed in the context of this set of studies. The analysis shows that ML and REML methods yield similar results.