The media's hand picked quotes from the thousands of documents/e-mails leaked from East Anglia suggest that there was pressure placed on researchers to do a little cosmetic work on graphs that did not perfectly support the thesis and to use data of questionable quality if it could be taken to support the thesis. The documented plans to destroy data rather than allow open evaluation is another issue, but regarding the pressure to "tweak" data, have any of you experienced it in your work?
I once had to take over a new material development project for a colleague who was leaving the company. The company's new material was supposed to provide increased resilience to rubber compounds. An analysis of the data my colleague had generated revealed that our new material actually acted as an inert filler, providing no measurable impact to any physical properties of the compound. My colleague had lead the R&D director to believe otherwise, and I had been given the data the afternoon before a big meeting with the customer we were wooing to fund further research. I showed the R&D director the analysis, and pointed out how the data he had been shown by my colleague was misleading. The director told me, "Keep quiet at the meeting tomorrow. The customer might not notice, and we can make a material that works if they fund the next phase of the project." I declined to attend the meeting, and two weeks later the customer came back with, "We were hoping for something with a little more activity," and severed their relationship with my company at the time since we provided "irreproducible results."
I've also watched a colleague spot trends in data where the signal-to-noise ratio hopelessly masked any trends that might be present. His confidence in his thesis lead to illusions in his perception. I know I've had to have holes in some of my data sets pointed out by more objective observers.
Have any of you ever been pressured to tweak or ignore data? How widespread do you think these sorts of things are in academia and industry?
Good Answers: