the inappropriate practice of searching through large files of information to try to confirm a preconceived hypothesis or belief without an adequate design that controls for possible confounds or alternate hypotheses. Data dredging may involve selecting which parts of a large data set to retain to get specific, desired results. An extreme example would be if a marketing researcher found that 91 out of 100 people surveyed were opposed to a certain product and then chose to focus only on the last 10 people in order to state that 9 out of 10 people prefer this product, when in fact there were only 9 out of 100 preferred it. See data snooping.