Evaluation of practical methods for estimating means for meta-analysis of continuous, non-normally-distributed outcomes