A principal investigator (PI) acquires data that portrays a wonderful correlation explaining a physical effect. About 10 percent of the data, however, lie far removed from the predicted values. There are explanations as to why these data are different. For example, some experimental parameters have not been well controlled and may have been different for the experiments in which the results deviate. The PI chooses to ignore the outlying data in the publication.

PLEASE CONSIDER THESE QUESTIONS

1. Should the PI have published the outlying data with an explanation of the limited generality of its correlation?

2. Should the PI have repeated the experiments for these data and ignored them only once they constituted less than five percent of the total data?

See Discussion

Source: University of Minnesota