A little pregnant.

There’s no such thing. Same goes for predefined statistical significance. Nevertheless, oncologists have a nasty habit of dropping qualifiers surrounding "borderline" significant (aka insignificant) p-values. The authors of this study painstakingly reviewed every original research article published in four major oncology journals from 2016 to 2017 to quantify “trend statements” describing “almost” significant results. Overall, 9% used these misleading statements to describe their results. More interesting was the trend (sorry) among journals. Nearly 15% of JCO articles “trended” results compared to 7% for JNCI and 4% for JAMA Onc. The Brits obviously have little tolerance for trend statements with <2% frequency in Lancet Oncology. That, or their trend statements were just too sophistically-worded to notice. TBL: Beware of statistically suggestive statements. | Nead, JAMA Oncol 2018


Popular Posts