Failure in Calculating The Statistics from Logarithm Values: Case Study

Abstract

       The statistics of an article on multi-center evaluation of the viral loads (Cytomegalo virus, CMV) presented in international journals (J Mol Diagn. Mar 2009; 11 (2): 87-92., were reviewed to check the validity of statistical calculation by using logarithmic rules. There were gross errors in log calculation of the means CMV loads which lead to missed conclusion that CMV loads amongst 23 laboratories were not different from the reference values, statistically. The Standard deviations (SD) was missed calculated from the log values directly, resulting in lower SD, compared to those calculated by using the absolute values. All statistics those using SD were miss-calculated : the pair t-test, 95% confidence intervals, Coefficient of Variance, Correlation Coefficient and affected the conclusions of the author that there was no difference in CMV load determination amongst these 23 laboratories. Hence, the significant CMV load variation amongst 23 laboratories were neglected and the need for reference materials were over-emphasized.