The consequences for being caught committing fraud (essentially termination of one’s career in most cases) are too high.
Not for “soft fraud” like data mining. And other types of fraud such as fudging the results of an experiment would be really hard to prove given that lots of honestly done experiments don’t seem to replicate. Having someone find an error in your analysis certainly isn’t cause for firing a tenured professor, and taking this into account I bet some people make deliberate errors that make their analysis more publishable. I’ve heard it can be sometimes very difficult to get another professor to give up his data, even when the data was used to publish an article in a journal that had a rule saying that you must make your data available upon request.
Not for “soft fraud” like data mining. And other types of fraud such as fudging the results of an experiment would be really hard to prove given that lots of honestly done experiments don’t seem to replicate. Having someone find an error in your analysis certainly isn’t cause for firing a tenured professor, and taking this into account I bet some people make deliberate errors that make their analysis more publishable. I’ve heard it can be sometimes very difficult to get another professor to give up his data, even when the data was used to publish an article in a journal that had a rule saying that you must make your data available upon request.