A researcher removes unusually high bills from a random sample before estimating the average monthly electric bill, but she did not have a rule for removing them before seeing the data. What is the main concern?