Objective To evaluate the right usage of statistical methods in medical articles. Methods 544 theses from eight medical journalspublished during1998 and 2005 were analyzed. Results 136 theses had obvious statistical errors,accounting for 25.00%. The main types of the errors were: the wrong methods of data analyzing for 61.76%, diagram error for 14.71%, nonstatistical dealing for8.82%, the mixture of rate and ratio for 8.82%, and the other error for 5.88%. Conclusion The statistical methods should be highly valued. Despite handing themanuscript to the specialist in the same field, the statistician checking system should be built through the process of manuscript dealing.