It is hard to estimate the prevalence of misconduct. Various surveys have tried to do this and produced different results probably because they used different methods and different definitions of misconduct. For example, asking researchers directly whether they have committed misconduct may not give accurate results. Similarly, we do not know how much misconduct goes undetected.
Daniele Fanelli has tried to answer this question by doing a systematic review of published studies (for which he screened 3207 papers). He combined results of 18 studies in a meta-analysis. He only included surveys of fabrication or falsification NOT plagiarism.
Estimate the percentage of researchers you think admitted to publication misconduct.