Hobbies and self-reported health. Data were first analyzed separately for each country using fixed effects regression. The findings were then pooled into multinational meta-analyses using the random effects model to estimate the overall effect sizes for all outcomes. Between-study heterogeneity was estimated using the algorithm of the restricted maximum likelihood and was assessed using I 2 and H 2 statistics. I 2 is the percentage of variability in the effect size that is caused by between-study heterogeneity, rather than by sampling error. The H 2 statistic describes the ratio of the observed variation and the expected variance due to sampling error. Given that some of the analyses had more participants than others and thus had lower sampling variability and more precise estimates, the meta-analysis was weighted. Studies with a greater number of respondents were given more weight than studies with a small number of respondents. These were relative weights that summed to 100. Data are presented as fixed effects coefficients and 95% CI. The overall effect size and its width should have accounted for the between-study variance, the number of studies, the precision of the study-specific estimates (or 'effect sizes') and the significance level.
