Number of studies, heterogeneity, generalisability, and the choice of method for meta-analysis