Effect sizes of fixed and random effects model in meta-analysis

Over a few decades ago, meta-analysis has rapidly developed in various fields of studies to improve the quality of previous research by quantitative review. Quantitative review assesses an overall magnitude of the effect from multiple studies. Hence, the objective in this dissertation is to determine the methods of finding effect sizes for different conditions in meta-analysis. Two key findings of this study are as follows: Firstly, vote-counting method will be applied if the effect size estimates are not available. Next, fixed effects and random effects methods will be used in combining the effect size estimates when effect size estimates can be calculated. 55 Randomised Controlled Trials (RCTs) on Tamoxifen treatment for early breast cancer patients are observed according to recurrences and mortality cases, respectively. The data is then analysed by using the latest statistical software, Comprehensive Meta-Analysis (CMA) version 2.0. CMA is found to be beneficial and interactive in the process of meta-analysis. The selection bias as well as the publication bias should be avoided in order to produce better quality meta-analysis.