A study looked at a random sample of 100 companies whose stocks were traded on the NYSE. The companies were divided into quintiles (the 20 largest companies, the 20 next largest, and so on) and the rates of return over the preceding 50 years were calculated for each quintile. This study found that the quintile with the smallest companies

(“small cap” stocks) had done better than the quintile with the largest companies (“large cap” stocks). Identify the most important statistical bias in this sample and explain how the study could be redone to get rid of this problem.