Fixed or random effects for meta-analysis: Theory and programming