This article is only available in the PDF format. Download the PDF to view the article, as well as its associated figures and tables.
To the Editor:—
In an article on the use of dramamine to control postoperative vomiting, in The Journal, Dec. 3, 1955, page 1342, the authors state that their statisticians at the University of Washington analyzed their results by the chi-square test and found p less than 1 in 100 million. From this, they say, the statisticians concluded that "the probability was less than one in a hundred million that the drug did not significantly reduce the incidence of postoperative vomiting." This conclusion is the result of a misleading and widespread error in the interpretation of medical statistics. What the test showed in this case is that, "On the assumption that the drug does not affect the incidence of postoperative vomiting, the probability is less than one in a hundred million that chi-square would have turned out as large as it did." But this is far different from the conclusion that
Zeisler EB. MEDICAL STATISTICS. JAMA. 1956;160(4):318. doi:10.1001/jama.1956.02960390068024