A case study on the effect of imputing the missing variability measures in meta analysis

A Meta Analysis is a statistical technique for integrating quantitative results of the same research question from several sources. It is being applied in various disciplines including the medical and social sciences research,. One of the common problems with meta analysis data is when the variability measures, specifically, the standard deviations are not reported in the individual studies. Current approaches in handling this problem is either to exclude the studies with the missing standard deviations or to impute the standard deviations. This paper examines and compares meta analysis estimates based on these two approaches using a meta analysis of all studies on the effectiveness of Transurethral Needle Ablation (TUNA) for treatment of Benign Prostatic Hyperplasia (BPH) patients. The results show that the estimates of treatment effects are more precise with inclusion of more studies using the imputed standard deviations. Additionally, the estimates are robust against different values of the correlation coefficients