Results 1 
4 of
4
From association to causation via regression
 Indiana: University of Notre Dame
, 1997
"... For nearly a century, investigators in the social sciences have used regression models to deduce causeandeffect relationships from patterns of association. Path models and automated search procedures are more recent developments. In my view, this enterprise has not been successful. The models tend ..."
Abstract

Cited by 31 (7 self)
 Add to MetaCart
(Show Context)
For nearly a century, investigators in the social sciences have used regression models to deduce causeandeffect relationships from patterns of association. Path models and automated search procedures are more recent developments. In my view, this enterprise has not been successful. The models tend to neglect the difficulties in establishing causal relations, and the mathematical complexities tend to obscure rather than clarify the assumptions on which the analysis is based. Formal statistical inference is, by its nature, conditional. If maintained hypotheses A, B, C,... hold, then H can be tested against the data. However, if A, B, C,... remain in doubt, so must inferences about H. Careful scrutiny of maintained hypotheses should therefore be a critical part of empirical work a principle honored more often in the breach than the observance.
Are There Algorithms That Discover Causal Structure? 30 June 1998
"... For nearly a century, investigators in the social and life sciences have used regression models to deduce causeandeffect relationships from patterns of association. Path models and automated search procedures are more recent developments. However, these formal procedures tend to neglect the diffic ..."
Abstract
 Add to MetaCart
For nearly a century, investigators in the social and life sciences have used regression models to deduce causeandeffect relationships from patterns of association. Path models and automated search procedures are more recent developments. However, these formal procedures tend to neglect the difficulties in establishing causal relations, and the mathematical complexities tend to obscure rather than clarify the assumptions on which the analysis is based. This paper focuses on statistical procedures that seem to convert association into causation. Formal statistical inference is, by its nature, conditional. If maintained hypotheses A, B, C,... hold, then H can be tested against the data. However, if A, B, C,...remain in doubt, so must inferences about H. Careful scrutiny of maintained hypotheses should therefore be a critical part of empirical work—a principle honored more often in the breach than the observance. Spirtes, Glymour, and Scheines have developed algorithms for causal discovery. We have been quite critical of their work. Korb and Wallace, as well as SGS, have tried to answer the criticisms. This paper will continue the discussion. Their responses may lead to progress in clarifying assumptions behind the methods, but there is little progress in demonstrating that the assumptions hold true for any real applications. The mathematical theory may be of some interest, but claims to have developed a rigorous engine for inferring causation from association are premature at best. The theorems have no implications for samples of any realistic size. Furthermore, examples used to illustrate the algorithms are diagnostic of failure rather than success. There remains a wide gap between association and causation. 1.
Are There Algorithms That Discover Causal Structure? 30 June 1998
"... For nearly a century, investigators in the social and life sciences have used regression models to deduce causeandeffect relationships from patterns of association. Path models and automated search procedures are more recent developments. However, these formal procedures tend to neglect the diffic ..."
Abstract
 Add to MetaCart
For nearly a century, investigators in the social and life sciences have used regression models to deduce causeandeffect relationships from patterns of association. Path models and automated search procedures are more recent developments. However, these formal procedures tend to neglect the difficulties in establishing causal relations, and the mathematical complexities tend to obscure rather than clarify the assumptions on which the analysis is based. This paper focuses on statistical procedures that seem to convert association into causation. Formal statistical inference is, by its nature, conditional. If maintained hypotheses A, B, C,... hold, then H can be tested against the data. However, ifA,B,C,...remain in doubt, so must inferences about H. Careful scrutiny of maintained hypotheses should therefore be a critical part of empirical work—a principle honored more often in the breach than the observance. Spirtes, Glymour, and Scheines have developed algorithms for causal discovery. We have been quite critical of their work. Korb and Wallace, as well as SGS, have tried to answer the criticisms. This paper will continue the discussion. Their responses may lead to progress in clarifying assumptions behind the methods, but there is little progress in demonstrating that the assumptions hold true for any real applications. The mathematical theory may be of some interest, but claims to have developed a rigorous engine for inferring causation from association are premature at best. The theorems have no implications for samples of any realistic size. Furthermore, examples used to illustrate the algorithms are diagnostic of failure rather than success. There remains a wide gap between association and causation. 1.
who have seen it, as one of the finest spectacles in Nature.
"... (SGS) is an ambitious book. SGS claim to have methods for discovering causal relations based only on empirical data, with no need for subjectmatter knowledge. These methods—which combine graph theory, statistics and computer science—are said to allow quick, virtually automated, conversion of associa ..."
Abstract
 Add to MetaCart
(SGS) is an ambitious book. SGS claim to have methods for discovering causal relations based only on empirical data, with no need for subjectmatter knowledge. These methods—which combine graph theory, statistics and computer science—are said to allow quick, virtually automated, conversion of association to causation. The algorithms are held out as superior to methods already in use in the social sciences (regression analysis, path models, factor analysis, hierarchical linear models, and so on). According to SGS, researchers who use these other methods are confused: Chapters 5 and 8 illustrate a variety of cases in which features of linear models that have been justified at length on theoretical grounds are produced immediately from empirical covariances by the procedures we describe.We also describe cases in which the algorithms produce plausible alternative models that show various conclusions in the social scientific literature to be unsupported by the data (p. 14). In the absence of very strong prior causal knowledge, multiple regression should not be used to select the variables that influence an outcome or criterion variable in data from uncontrolled studies. So far as we can tell, the popular automatic regression search procedures [like stepwise regression] should not be used at all in contexts where causal inferences are at stake. Such contexts require improved versions of algorithms like those described here to select those variables whose influence on an outcome can be reliably estimated by regression