What is the minimum number of effect sizes required in meta-regression? An estimation based on statistical power and estimation precision