Prevalence of data fraud

I thank Professor Wiwanitkit for his letter. The primary purpose of my paper [1], as the title suggests, was to review the literature on the prevalence of scientific misconduct and contributing factors leading to misconduct. Although the existing evidence on these topics is rather problematic as I noted in the paper “The evidence on prevalence is unreliable and fraught with definitional problems and with study design issues”, I believe a reasonable interpretation of the available data indeed suggests that “…cases of the most serious types of misconduct, fabrication and falsification (i.e., data fraud), are relatively rare…”. Of course, the definition of ‘rare’ is not clear and it can be argued that an overall prevalence of 2 %, as suggested by a meta-analysis of studies of self-reported admissions of fabrication or falsification [2], not only suffers from a potentially large under-reporting bias but, even taken at face value, indicates a situation that is in fact not rare for such a serious breach of scientific ethics. In any case, data fraud does occur regularly, rarely or not, and it is certainly true that the complex issues of prevention and detection of data fraud deserve serious attention.