From 2007, there has been a huge proliferation in the discovery of genetic variants affecting human traits and diseases, achieved largely by the integration of multiple genome-wide association studies (GWAS) via meta-analysis. The principal objective of this thesis is to develop robust approaches for meta-analysis GWAS in order to reduce false positive findings and optimise statistical power. I consider both Single Nucleotide Polymorphism (SNP) and Copy Number Variant (CNV) GWAS. First, to gain background knowledge in GWAS and meta-analysis, I was involved in a large-scale meta-analysis GWAS to identify genetic variants associated with alcohol consumption, as the main statistical analyst. This study provided me with the opportunity to investigate ways of reducing the probability of false positive findings, via quality control procedures. The main discovery from the study was the identification of the Autism susceptibility candidate 2 gene (AUTS2) as associated with alcohol consumption at genome-wide significance. In the alcohol study, different phenotype transformations were applied to the data according to the inclusion or exclusion of non-drinkers, which led to questioning which transformation of skewed continuous phenotypes optimises statistical power in GWAS in general, forming the second major investigation in my thesis. It was shown that while the inverse normal transformation (INT) may not be the preferable choice of transformation in many epidemiological studies where effect sizes are large, its application to non-normal phenotypes in GWAS, where effect sizes are small and the priority is discovery over interpretability, may lead to an increase in the discovery of genetic variants affecting continuous traits. Finally, as knowledge about CNVs has accumulated in recent years, the meta-analysis of GWAS on CNVs has become a natural next step forward in the field. Therefore, I investigated and developed an approach to enable CNV meta-analysis to proceed in a similar way as SNP meta-analyses. This approach was developed into a software package, cnvPipe, which was applied to investigate CNVs associated with height and weight in the meta-analysis setting.