Premium
TWO MODEL‐BASED INFERENCE ARGUMENTS IN SURVEY SAMPLING
Author(s) -
Särndal C. E.
Publication year - 1980
Publication title -
australian journal of statistics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.434
H-Index - 41
eISSN - 1467-842X
pISSN - 0004-9581
DOI - 10.1111/j.1467-842x.1980.tb01184.x
Subject(s) - inference , sampling (signal processing) , fiducial inference , simple (philosophy) , bayesian inference , computer science , point (geometry) , bayesian probability , point estimation , mathematics , econometrics , statistics , frequentist inference , artificial intelligence , epistemology , philosophy , geometry , filter (signal processing) , computer vision
Summary Which of the two model‐based inference arguments in survey sampling is the “correct” one? The answer seems simple once the goal of the estimation procedure has been clearly defined, but the literature to date on the topic is inconclusive and likely to mislead sampling practitioners. On the one hand we have the approach used by a classical exponent of sampling theory such as Cochran (1977) in those rare instances where he does invoke a model, on the other hand we have the prediction approach formalized by Brewer (1963) and Roy all (1970). (The latter is also available in Bayesian and fiducial versions.) Both approaches are valid, but it is necessary to make a distinction between them. The results they yield are not always identical. In this paper we describe the differences between the two approaches and point to the conditions under which they agree or disagree.