Reported by a reader on May 31, 2004
Should Research Results be Trusted?
(Ivanhoe Newswire) – You might think basing medical practice on published evidence sounds like a good idea, but what if the published findings are inaccurate? A new study shows a large proportion of articles in top science and medical journals contain statistical errors.
According to research published this week in BMC Medical Research Methodology, 38 percent of Nature papers and 25 percent of British Medical Journal articles that were evaluated contained at least one statistical error. In total, more than 11 percent of the statistical results published were incongruent, and 4 percent of these errors may have caused non-significant findings to be misrepresented as being significant.
The researchers hypothesize that the errors are due to inaccurate rounding or transcription. For example, study authors might round up 2.38 to 2.5 rather than 2.4 to make the numbers look neater.
These results were reached upon checking statistical study results in 181 papers of the 2001 volumes of Nature and in 63 random articles of the 2001 volumes of the British Medical Journal.
“Although these kinds of errors may leave the conclusions of a study unchanged, they are indicative of poor practice,” the researchers comment. “Our concern is that these kinds of errors are probably present in all numerical results and all steps of scientific research, with potentially important practical consequences.”
One way to minimize the effect of these errors, researchers suggest, would be for published authors to make their raw data freely available on the Internet. They say this would allow other researchers to check the results and conclusions of the study for themselves, making fraud and sloppiness more easily detected.
SOURCE: BMC Medical Research Methodology, 2004;4:13
Incongruence between test statistics and P values in medical papers
Emili García-Berthou and Carles Alcaraz
Department of Environmental Sciences, University of Girona, E-17071 Girona, Spain
BMC Medical Research Methodology 2004, 4:13