Correction

Abstract

CORRECTION We would like to report an erratum to the paper 'What to add to nothing? Use and avoidance of continuity corrections in meta-analysis of sparse data', Statistics in Medicine 2004; 23:1351-1375. The error aaects the calculation of the bias and coverage statistics for the Mantel-Haenszel estimator when no continuity correction is applied (Table X and Figure 4). For a 1:1, 1:2 and 1:4 group imbalance the correct absolute biases are −0:01, −0:04 and −0:06, respectively (not −0:20, −0:10 and −0:08). For a 1:1 and 1:4 group imbalance the correct coverages are 96.7 per cent and 95.0 per cent (not 94.6 per cent and 95.1 per cent). Hence the Mantel-Haenszel estimator without continuity correction performs well when there is no group imbalance in a sparse data situation. All other results and conclusions remain unchanged. We deeply regret any inconvenience caused by this computational error.