Cournot sobre el "efecto Roseto", 120 años antes de tal
Esta entrada abunda sobre una de la semana pasada sobre el llamado efecto Roseto. El Cournot al que alude el titulo es el Cournot famoso (1801-1877) al que, a pesar de ser más conocido por sus aportaciones a la economía, debemos una Exposition de la théorie des chances et des probabilités de 1843.
En su párrafo 114 critica explícitamente el tipo de conclusiones a las que llegan los descuidados exégetas del asunto Roseto y que Stigler comenta así:
Cournot’s rejection of the a posteriori meaningfulness of probability was not based on reservations about a priori assumptions on the unknown chances $latex x_1$ and $latex x_2$ such as might occur to a modern critic, but rather upon an acute awareness of the even more devastating effect that selecting a hypothesis after the data is at hand could have upon the conclusion. For emphasis, he supposed that one of the eighty-six departments of France exhibited a much larger ratio of male births than did all of France treated as a whole. For the corresponding a posteriori probability $latex \Pi$ to have an objective meaning, Cournot argued (pp. 197 - 198), it would be necessary that we know that the department in question was selected at random, as if its name were drawn from among eighty-six tickets in an urn and not merely singled out as the department with the largest male birth ratio among the eighty-six. But if the departments were selected at random or specified by the statistician prior to the analysis of the data, Cournot would have no difficulty accepting the a posteriori probability $latex \Pi$.