Clark Glymour, Richard Scheines, Peter Spirtes and Kevin Kelly. Discovering Causal Structure: Artifical Intelligence, Philosophy of Science and Statistical Modeling.

Causal modeling methods such as path analysis, used in the social and natural sciences, are also highly relevant to philosophical problems of probabilistic causation and statistical explanation. We show how these methods can be effectively used (1) to improve and extend Salmon's S-R basis for statistical explanation, and (2) to repair Cartwright's resolution of Simpson's paradox, clarifying the relationship between statistical and causal claims.

Recent philosophical studies of probabilistic causation and statistical explanation have opened up the possibility of unifying philosophical approaches with causal modeling as practiced in the social and biological sciences. This unification rests upon the statistical tools employed, the principle of common cause, the irreducibility of causation to statistics, and the idea of causal process as a suitable framework for understanding causal relationships. These four areas of contact are discussed with emphasis on the relevant aspects of causal modeling.

(2013). Statistical models as cognitive models of individual differences in reasoning. Argument & Computation: Vol. 4, Formal Models of Reasoning in Cognitive Psychology, pp. 89-102. doi: 10.1080/19462166.2012.674061.

Two Bayesian approaches to choosing between statistical models are contrasted. One of these is an approach which Bayesian statisticians regularly use for motivating the use of AIC, BIC, and other similar model selection criteria, and the other one is a new approach which has recently been proposed by Bandyopadhayay, Boik, and Basu. The latter approach is criticized, and the basic ideas of the former approach are presented in a way that makes them accessible to a philosophical audience. It is (...) also pointed out that the former approach establishes a new, philosophically interesting connection between the notions of simplicity and informativeness. (shrink)

Determining optimal units of representing morphologically complex words in the mental lexicon is a central question in psycholinguistics. Here, we utilize advances in computational sciences to study human morphological processing using statistical models of morphology, particularly the unsupervised Morfessor model that works on the principle of optimization. The aim was to see what kind of model structure corresponds best to human word recognition costs for multimorphemic Finnish nouns: a model incorporating units resembling linguistically defined morphemes, a whole-word model, or (...) a model that seeks for an optimal balance between these two extremes. Our results showed that human word recognition was predicted best by a combination of two models: a model that decomposes words at some morpheme boundaries while keeping others unsegmented and a whole-word model. The results support dual-route models that assume that both decomposed and full-form representations are utilized to optimally process complex words within the mental lexicon. (shrink)

I discuss the philosophical significance of the statistical model selection criteria, in particular their relevance for philosophical of underdetermination. I present an easily comprehensible account of their simplest possible application and contrast it with their application to curve-fitting problems. I embed philosophers' earlier discussion concerning the situations in which the criteria yield implausible results into a more general framework. Among other things, I discuss a difficulty which is related to the so-called subfamily problem, and I show that it has (...) analogies in all legitimate applications of the model selection criteria, and that an analogy of Goodman's new riddle of induction can be formulated in only some of their applications. (shrink)

Selectional preferences have a long history in both generative and computational linguistics. However, since the publication of Resnik's dissertation in 1993, a new approach has surfaced in the computational linguistics community. This new line of research combines knowledge represented in a pre‐defined semantic class hierarchy with statistical tools including information theory, statistical modeling, and Bayesian inference. These tools are used to learn selectional preferences from examples in a corpus. Instead of simple sets of semantic classes, selectional preferences are (...) viewed as probability distributions over various entities. We survey research that extends Resnik's initial work, discuss the strengths and weaknesses of each approach, and show how they together form a cohesive line of research. (shrink)

This Element has two main aims. The first one is an historically informed review of the philosophy of probability. It describes recent historiography, lays out the distinction between subjective and objective notions, and concludes by applying the historical lessons to the main interpretations of probability. The second aim focuses entirely on objective probability, and advances a number of novel theses regarding its role in scientific practice. A distinction is drawn between traditional attempts to interpret chance, and a novel methodological study (...) of its application. A radical form of pluralism is then introduced, advocating a tripartite distinction between propensities, probabilities and frequencies. Finally, a distinction is drawn between two different applications of chance in statistical modelling which, it is argued, vindicates the overall methodological approach. The ensuing conception of objective probability in practice is the 'complex nexus of chance'. (shrink)

A model of causation is presented which shares the advantages of Reichenbach's definition in terms of the screening-off relation, but which has the added advantage of distinguishing cause and effect without reference to temporal directionality. This model is defined in terms of the masking relation, which in turn is defined in terms of the equivocation relation of communication theory.

A fundamental question in visual neuroscience is: Why are the response properties of visual neurons as they are? A modern approach to this problem emphasizes the importance of adaptation to ecologically valid input, and it proceeds by modeling statistical regularities in ecologically valid visual input (natural images). A seminal model was linear sparse coding, which is equivalent to independent component analysis (ICA), and provided a very good description of the receptive fields of simple cells. Further models based on modeling (...) residual dependencies of the ‘‘independent” components have later been introduced. These models lead to emergence of further properties of visual neurons: the complex cell receptive fields, the spatial organization of the cells, and some surround suppression and Gestalt effects. So far, these models have concentrated on the response properties of neurons, but they hold great potential to model various forms of inference and learning. (shrink)

The question of whether it is possible to characterise grammatical knowledge in probabilistic terms is central to determining the relationship of linguistic representation to other cognitive domains. We present a statistical model of grammaticality which maps the probabilities of a statistical model for sentences in parts of the British National Corpus (BNC) into grammaticality scores, using various functions of the parameters of the model. We test this approach with a classifier on test sets containing different levels of syntactic (...) infelicity. With appropriate tuning, the classifiers achieve encouraging levels of accuracy. These experiments suggest that it may be possible to characterise grammaticality judgements in probabilistic terms using an enriched language model. (shrink)

This paper shows how to define probability distributions over linguistically realistic syntactic structures in a way that permits us to define language learning and language comprehension as statistical problems. We demonstrate our approach using lexical‐functional grammar (LFG), but our approach generalizes to virtually any linguistic theory. Our probabilistic models are maximum entropy models. In this paper we concentrate on statistical inference procedures for learning the parameters that define these probability distributions. We point out some of the practical problems (...) that make straightforward ways of estimating these distributions infeasible, and develop a “pseudo‐likelihood” estimation procedure that overcomes some of these problems. This method raises interesting questions concerning the nature of the data available to a language learner and the modularity of language learning and processing. (shrink)

The growing range of methods for statistical model selection is inspiring new debates about how to handle the potential for conflicting results when different methods are applied to the same data. While many factors enter into choosing a model selection method, we focus on the implications of disagreements among scientists about whether, and in what sense, the true probability distribution is included in the candidate set of models. While this question can be addressed empirically, the data often provide inconclusive (...) results in practice. In such cases, we argue that differences in prior metaphysical views about the local adequacy of the models can produce underdetermination of results, even for the same data and candidate models. As a result, data alone are sometimes insufficient to settle rational beliefs about nature. (shrink)

I discuss the philosophical significance of the statistical model selection criteria, in particular their relevance for philosophical problems of underdetermination. I present an easily comprehensible account of their simplest possible application and contrast it with their application to curve‐fitting problems. I embed philosophers' earlier discussion concerning the situations in which the criteria yield implausible results into a more general framework. Among other things, I discuss a difficulty which is related to the so‐called subfamily problem, and I show that it (...) has analogies in all legitimate applications of the model selection criteria, and that an analogy of Goodman's new riddle of induction can be formulated in only some of their applications. (shrink)

There is considerable confusion about the role of p-values in statistical model checking. To clarify that point, I introduce the distinction between measures of surprise and measures of evidence which come with different epistemological functions. I argue that p-values, often understood as measures of evidence against a null model, do not count as proper measures of evidence and are closer to measures of surprise. Finally, I sketch how the problem of old evidence may be tackled by acknowledging the epistemic (...) role of surprise indices. (shrink)

We give an account of the appearance and first developments of the statistical model of atoms proposed by Thomas and Fermi, focusing on the main results achieved by Fermi and his group in Rome. Particular attention is addressed to the unknown contribution to this subject by Majorana, anticipating some important results reached later by leading physicists.

Doctors’ survival predictions for terminally ill patients have been shown to be inaccurate and there has been an argument for less guesswork and more use of carefully constructed statistical indices. As statisticians, the authors are less confident in the predictive value of statistical models and indices for individual survival times. This paper discusses and illustrates a variety of measures which can be used to summarise predictive information available from a statistical model. The authors argue that models and (...) statistical indices can be useful at the group or population level, but that human survival is so uncertain that even the best statistical analysis cannot provide single-number predictions of real use for individual patients. (shrink)

In a recent article [1], Kenneth Sayre presents what he takes to be an alternative to Salmon and Reichenbach's screening-off model of causal relations. His statistical model, based on communication theory, supposedly “has the added advantage of distinguishing cause and effect without reference to temporal distinction”. Unfortunately his model falls far short of this grand claim, faltering before the same sort of examples I had previously used to assail the screening-off model [2].

Classification in statistics is usually used to solve the problems of identifying to which set of categories, such as subpopulations, new observation belongs, based on a training set of data containing information whose category membership is known. The article aims to use the Gaussian Mixture Model to model the daily closing price index over the period of 1/1/2013 to 16/8/2020 in the Kingdom of Saudi Arabia. The daily closing price index over the period declined, which might be the effect of (...) corona virus, and the mean of the study period is about 7866.965. The closing price is the last regular deal that took place during the continuous trading period. If there are no transactions on the stock during the day, the closing price is the previous day’s closing price. The closing auction period comes after the continuous trading period, during which investors can enter by buying and selling the stocks at this period. The experimental results show that the best mixture model is E with three components according to the BIC criterion. The expectation-maximization algorithm converged in 2 repetitions. The data source is from Tadawul KSA. (shrink)

Drawing substantive conclusions from linear causal models that perform acceptably on statistical tests is unreasonable if it is not known how alternatives fare on these same tests. We describe a computer program, TETRAD, that helps to search rapidly for plausible alternatives to a given causal structure. The program is based on principles from statistics, graph theory, philosophy of science, and artificial intelligence. We describe these principles, discuss how TETRAD employs them, and argue that these principles make TETRAD an effective (...) tool. Finally, we illustrate TETRAD's effectiveness by applying it to a multiple indicator model of Political and Industrial development. A pilot version of the TETRAD program is described in this paper. The current version is described in our forthcoming Discovering Causal Structure: Artificial Intelligence for Statistical Modeling. (shrink)

In a recent paper, Sprenger (2019) advances what he calls a “suppositional” answer to the question of why a Bayesian agent’s degrees of belief should align with the probabilities found in statistical models. We show that Sprenger’s account trades on an ambiguity between hypothetical and subjunctive suppositions and cannot succeed once we distinguish between the two.

Statistical model specification and validation raise crucial foundational problems whose pertinent resolution holds the key to learning from data by securing the reliability of frequentist inference. The paper questions the judiciousness of several current practices, including the theory-driven approach, and the Akaike-type model selection procedures, arguing that they often lead to unreliable inferences. This is primarily due to the fact that goodness-of-fit/prediction measures and other substantive and pragmatic criteria are of questionable value when the estimated model is statistically misspecified. (...) Foisting one's favorite model on the data often yields estimated models which are both statistically and substantively misspecified, but one has no way to delineate between the two sources of error and apportion blame. The paper argues that the error statistical approach can address this Duhemian ambiguity by distinguishing between statistical and substantive premises and viewing empirical modeling in a piecemeal way with a view to delineate the various issues more effectively. It is also argued that Hendry's general to specific procedures does a much better job in model selection than the theory-driven and the Akaike-type procedures primary because of its error statistical underpinnings. (shrink)

It is probable that, increasingly, genome investigations are going to be based on statistical formalization. This review summarizes the state of art and potentiality of using statistics in microbial genome analysis. First, I focus on recent advances in functional genomics, such as finding genes and operons, identifying gene conversion events, detecting DNA replication origins and analysing regulatory sites. Then I describe how to use phylogenetic methods in genome analysis and methods for genome‐wide scanning for positively selected amino acids. I (...) conclude with speculations on the future course of genome statistical modeling. BioEssays 25:266–273, 2003. © 2003 Wiley Periodicals, Inc. (shrink)

Over the past century, private tutoring in many countries has increased dramatically. Moreover, the main disadvantage of PT is that has a byproduct and a characteristic on the educational system in developing countries in terms of contributing to conditions such as large class sizes, low public expenditures, and an inadequate number of universities. In Saudi Arabia, the spread of PT at school and university levels has yet to be addressed by researchers. One goal of this examination was to research the (...) spread of PT in relation to mathematics being taught at school and university levels. This investigation used quantitative data of the questionnaire to determine the reasons for the spread. A total of 1000 students responded from University of Ha’il. The edge plans given results that are associated and like the assessment that is performed with the whole data. The current study found that a good family financial situation and lack of parental supervision of the children were significant factors underlying the spread of private tutoring at school and university levels. (shrink)

This article investigates a problem for statistical model evaluation, in particular for curve fitting: by employing a different family of curves we can fit any scatter plot almost perfectly at apparently minor cost in terms of model complexity. The problem is resolved by an appeal to prior probabilities. This leads to some general lessons about how to approach model evaluation.

Yarkoni recommends that psychology researchers should take care to align their statistical models to the verbal theories they are studying and testing. This principle applies not just to qualitative theories in psychology but also to more quantitative sciences: there, too, mismatch between open-ended theories and specific statistical models have led to confusion.