THEY may be two of the world鈥檚 top scientific journals, but Nature and the British Medical Journal have been found guilty of routinely publishing figures that don鈥檛 add up. 鈥淚t鈥檚 another example that statistical practice is generally poor, even in the most renowned scientific journals,鈥� says Emili Garc铆a-Berthou and Carles Alcaraz of the University of Girona in Spain.
The pair analysed 32 papers published in Nature in 2001 and 12 in the BMJ. One or more statistical errors appeared in 38 per cent of Nature papers and 25 per cent of BMJ papers. And up to 4 per cent of statistically 鈥渟ignificant鈥� results reported in these papers may not be, they report in BMC Medical Research Methodology (vol 4, article 13). The pair say raw data should be published online for other researchers to check.