Per Gösta Andersson

Is there anything more fundamental within the area of statistical inference than the estimation, in this case interval estimation, of a probability based on success/failure trials? The problem is conceptually easy for students to understand, but finding a satisfactory solution turns out to be an exercise that is far from trivial. Moreover, even learned scholars disagree as to which procedure to present in a teaching situation. Among several suggestions in the literature over the years, we will concentrate on Wald and Wilson intervals. From recent comments by colleagues, it seems there is still a believe that the Wald interval is worth bringing up as something useful, at least for large samples, in spite of all its deficiencies. What is most worrying is that, even today, a considerable amount of well-established and influential textbooks actually advocate the Wald interval.

Some content is only viewable by ASA Members. Please login or become an ASA member to gain access.