Linear Mixed Model Effects in Meta Analysis of Agricultural Data

This paper proposes a suitable approach to analyze agricultural data by fitting a linear mixed model in a metaanalysis (MA) framework. We develop an underlying theory that provides practical information on how to carry out the analysis in the MA by using the Maximum Likelihood Estimate (MLE) method to estimate the parameters in the model without requiring an assumption about a specific form of the population distribution. Past studies in the MA assumed of independence and normality of the data, but this is not always the case. The concept of log-likelihood approach enables us to estimate the parameters of interests through the process of log-odds ratio in the MA. There is a need for a mechanism of testing the past studies to find a good estimate of a parameter of interest in the absence of assumptions. This mechanism of transformation in the analysis is based on the accuracy of the estimate parameter from the dataset. An illustrative example on maize grain yield is used to determine whether the efficacy of using fertiliser given that the irrigation was used in the experiment to compare to the efficacy of fertiliser application under the rainfall.. KEYWORDS :Fixed effect model, Random effect model, Log-Odds ratio, Meta analysis, Maximum likelihood estimate.