Statisticians are often asked by researchers to 'massage' data to get the results they want, a new study has found according to Bloomberg.
New York University's Ralph Katz, an epidemiologist, tells the CBC that he got the idea for his survey after chatting with a statistician and asking how often statisticians are asked by researchers to manipulate data. "I just asked him one day when we were chatting. And he said 'frequently,'" he says.
Katz and his colleagues surveyed 522 members of the American Statistical Association who identified as consulting biostatisticians to inquire how often they are asked to stress significant findings over non-significant ones, falsify statistical significance, or not report missing data, and more. As they reported recently in the Annals of Internal Medicine, the researchers found statisticians were often asked to do such things like interpret the findings based on expectations or removed some data records to better fit the hypothesis.
A related editorial in the Annals says this could contribute to the issue of nonreproducible research.
Bloomberg's Faye Flam also notes that researchers have incentives to have bright, positive results. "From scientists' point of view, a study is a failure if it gets an answer that doesn't help their careers," she writes. "For the rest of us, a failed study is one whose results don't correspond to the truth."