1 in 7 Scientists Say Colleagues Fake Data

[Winston]

http://www.timesonline.co.uk/tol/news/u ... 425036.ece

Faking scientific data and failing to report commercial conflicts of
interest are far more prevalent than previously thought, a study
suggests.

One in seven scientists says that they are aware of colleagues having
seriously breached acceptable conduct by inventing results. And around
46 per cent say that they have observed fellow scientists engage in
"questionable practices", such as presenting data selectively or
changing the conclusions of a study in response to pressure from a
funding source.

However, when scientists were asked about their own behaviour only 2
per cent admitted to having faked results.

Daniele Fanelli, of the University of Edinburgh, who carried out the
investigation, believes that high-profile cases such as that of Hwang
Woo-Suk, the South Korean scientist disgraced for fabricating human
stem cell data, are less unusual than is generally assumed.
"Increasing evidence suggests that known frauds are just the tip of
the iceberg and that many cases are never discovered," he said.

The findings, published in the peer-reviewed journal PLoS One, are
based on a review of 21 scientific misconduct surveys carried out
between 1986 and 2005. The results paint a picture of a profession in
which dishonesty and misrepresentation are widespread.

In all the surveys people were asked about both their own research
practices and those of colleagues. Misconduct was divided into two
categories: fabrication, the actual invention of data; and lesser
breaches that went under the heading "questionable practices". These
included dropping data points based on a "gut feeling" and failing to
publish data that contradict one's previous research.

The discrepancy between the number of scientists owning up to
misconduct and those having been observed by colleagues is likely to
be in part due to fears over anonymity, Dr Fanelli suggests. "Anyone
who has ever falsified research is probably unwilling to reveal it
despite all guarantees of anonymity."

The study predicts that the 2 per cent figure, although higher than
most previous estimates, is still likely to be conservative.

Another explanation for the differences between the self-report
results and colleague-report results could be that people consider
themselves to be more moral than others. In a marginal case, people
might characterise their colleagues' behaviour as misconduct more
readily than they would their own.

The study included scientists from a range of disciplines. Misconduct
was far more frequently admitted by medical or pharmacological
researchers than others, supporting fears that the field of medical
research is being biased by commercial interests.

[Master]

wow. is everything a big fat lie?

[Winston]

More interesting info has come out about why published findings by the science establishment can't be totally trusted.

This one from the National Institute of Health library.

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1182327/

Why Most Published Research Findings Are False

Summary
There is increasing concern that most current published research findings are false. The probability that a research claim is true may depend on study power and bias, the number of other studies on the same question, and, importantly, the ratio of true to no relationships among the relationships probed in each scientific field. In this framework, a research finding is less likely to be true when the studies conducted in a field are smaller; when effect sizes are smaller; when there is a greater number and lesser preselection of tested relationships; where there is greater flexibility in designs, definitions, outcomes, and analytical modes; when there is greater financial and other interest and prejudice; and when more teams are involved in a scientific field in chase of statistical significance. Simulations show that for most study designs and settings, it is more likely for a research claim to be false than true. Moreover, for many current scientific fields, claimed research findings may often be simply accurate measures of the prevailing bias. In this essay, I discuss the implications of these problems for the conduct and interpretation of research.

https://www.theatlantic.com/magazine/ar ... ce/308269/

Lies, Damned Lies, and Medical Science
Much of what medical researchers conclude in their studies is misleading, exaggerated, or flat-out wrong. So why are doctors—to a striking extent—still drawing upon misinformation in their everyday practice? Dr. John Ioannidis has spent his career challenging his peers by exposing their bad science.

https://www.statnews.com/2017/08/02/ran ... -research/

Randomized controlled trials have long been held up as the “gold standard” of clinical research. There’s no doubt that well-designed trials are effective tools for testing a new drug, device, or other intervention. Yet much of modern medical care — perhaps most of it — is not based on randomized controlled trials and likely never will be. In this “dark matter” of clinical medicine, past practices and anecdotes all too often rule. We need to look beyond trials to improve medical care in these areas.

[Taco]

[Moretorque]

Which means depopulation...

[MrMan]

In the hard sciences like physics, numerous scientists repeat the same experience and co-publish the data, so that decreases the risk. I think some of the medical research may be similar. But there are a lot of science and social science papers that are not like this. Maybe more names on the paper makes it less likely data is faked, but a name could be added on the paper for helping with the literature review.

Academics have to produce papers to keep their jobs, to stay hireable, to be able to move up. They could spend a year collecting data and get not statistically significant results, so I understand the temptation. I've never faked data. But there are things you can do to legitimately take the same data and get statistically different results. I heard an example of someone studying rainfall. Change in rainfall was not statistically significant, but percentage of the change in rainfall was. The professor presenting this information complained about how data mining went against the assumptions of hypothesis testing using conventional statistics, but not Bayesian statistics.