6.43. Ideally, results of a statistical analysis should not depend greatly on a single observation. To check this, it's a good idea to conduct a sensitivity study:

Redo the analysis after deleting an outlier from the data set or changing its value to a more typical value, and check whether results change much. For the anorexia data shown in Example 5.5

(page 120), the weight change of 20.9 pounds was a severe outlier. Suppose this observation was actually 2.9 pounds but was incorrectly recorded. Redo the one-sided test of Example 6.4 (page 151), and summarize the influence of that observation.