We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
STATISTICAL INFERENCE IN CAUSAL ANALYSIS: SOME FOUNDATIONS.
- Authors
Birnbaum, Ian
- Abstract
The analysis of recursive systems incorporates statistical inferential methods both for the estimation of parameters and for the testing of specific causal models. This holds whether least-squares or maximum likelihood methods are applied. In the former case, the process is distribution free but when distributional assumptions are lacking formal model testing is prohibited. Mathematician K. Popper argues that probabilities should be viewed as weights which are dispositional properties of the conditions, that is to say propensities. Thus the conditions responsible for generating the possible values of a variable have a propensity to produce each value to a degree recorded by the weight or probability associated with the value. This formulation allows to interpret the probability of a singular event as a property of the singular event itself. Popper then defines the generating conditions as the conditions people intend to keep constant during repetition and so undermines the application of his interpretation of probability to the singular case. It is important to realize, however, that any inferences of a statistical nature drawn as to the values of the parameters of the causal model refer only to the data under analysis. There is no justification for using statistical inference to go beyond this data given the basis upon which such inference has been validated here.
- Subjects
MATHEMATICAL statistics; PARAMETERS (Statistics); POPPER, Karl Raimund, Sir, 1902-1994; PROBABILITY theory; DATA analysis; STATISTICAL correlation
- Publication
Quality & Quantity, 1979, Vol 13, Issue 3, p203
- ISSN
0033-5177
- Publication type
Article
- DOI
10.1007/BF00170037