Talk:A priori (statistics)

Is this term really used by statisticians? I've always heard the term "prior" rather than a priori, and the latter term obviously confuses "prior" with a term used in epistemology with a clearly different meaning. I know "a priori probability" is written about all the time by non-statisticians. I regard that as an incorrect usage by confused people. Michael Hardy (talk) 03:44, 22 April 2012 (UTC)
 * The term is used rather than "prior" by statisticians with a poor command of English (or impoverished language, e.g. Swedish) or malevolence; "antecedent" would be better for "prior". I think that the "Prior Analytics" discusses a priori propositions.
 * The link to the Mood textbook may be a pirated .pdf copy, to which the link should be removed, per WP:NEVER. Kiefer .Wolfowitz 08:04, 22 April 2012 (UTC)


 * The term has been used and seems still to be used ... but the question is really whether there is a special usage in statistics. Statistical theory sources contain mention of "a priori hypothesis" (emphasizing the "before data" status), and there may also be use in analysis of factorial experiments where there is an a priori assumption that certain higher order interactions don't exist. These uses don't relate directly to assumed probabilities, but to assumed model structures. Some sources found by google scholar are:


 * Maurer W, Hothorn LA, Lehmacher W. Multiple comparison in drug clinical trials and preclinic assays: a-priori ordered hypotheses. In: Vollmer J, editor. Biometrie in der pharmazeutischen Industrie. German. Volume 6. Testing principles in clinical and preclinical trials. Stuttgart: Fischer; 1995. p 3-18.


 * Kullback, S.; Leibler, R. A. (1951). On information and sufficiency. Annals of Mathematical Statistics, 22, 79–86.




 * There are also uses of a priori in contrast to post hoc. But in most cases the use of a priori seems to be covered by its more general use in the English language. Melcombe (talk) 12:23, 23 April 2012 (UTC)