Doll predetermined the distribution of the response variable ( this is done in case- control studies). Explain why each of the proportions in Question 28 cannot appropriately be extended to a larger population. Is it appropriate to conclude that 60 / 120 = 50% of female patients in the 20 London hospitals have lung cancer? Is it appropriate to assume that a good estimate of the percentage of female nonsmoking patients in the hospitals who have lung cancer is 37.2%? That tests for equal proportions are not appropriate if the response totals are fixed prior to the study. In addition, the next section will show that since the response totals are not random, a test of independence should not be used. Fisher’s exact test and a simulation study could be used. The following questions will show why it is appropriate to conduct a hypothesis test about the odds ratio.