Statistical Testing in Industrial Economics: A Reply on Measurement Error and Sampling Procedure

Abstract

PROFESSOR Telser's opening paragraphs betray a curiously scholastic view of research. He sees our original paper as an attempt to refute his earlier work on the association between concentration and advertising intensity. Finding that our 'badly biased' sample '. . . seriously distorts the true state of affairs', he promises to test 'fairly' the hypothesis at issue. We believe that the term 'refute' is more suitable to a discussion involving issues of deductive rather than inductive inference. It is inappropriate as a synonym for 'disagree with', especially when the disagreement is one of degree. What we did, of course, was simply to uncover a statistical relationship between concentration and advertising intensity that was stronger than that uncovered earlier by Telser. It may be that Telser has tested the hypothesis fairly, though we shall present some comments on this below. But it is not clear to us how Telser knows what the true state of affairs is. Further, since Telser nowhere even attempts to demonstrate bias in a statistical sense, we wonder whether his conclusion that our sample is badly biased might have been developed from a chain of reasoning somewhat as follows. If one were in a position to know that the true correlation between concentration and advertising intensity was zero, then he might be tempted to conclude that any sample yielding a strong non-zero correlation must be badly biased. We turn now to the specific issues raised by Telser.'