ABSTRACTWe investigate the applicability of Rodrik’s accounts of model selection and horizontal progress to macroeconomic DSGE modelling in both academic and policy-oriented modelling contexts. We argue that the key step of identifying critical assumptions is complicated by the interconnectedness of the common structural core of DSGE models and by the ad hoc modifications introduced to model various rigidities and other market imperfections. We then outline alternative ways in which macroeconomic modelling could become more horizontally progressive.

Several quantitative techniques for choosing among data models are available. Among these are techniques based on algorithmic information theory, minimum description length theory, and the Akaike information criterion. All these techniques are designed to identify a single model of a data set as being the closest to the truth. I argue, using examples, that many data sets in science show multiple patterns, providing evidence for multiple phenomena. For any such data set, there is more than one data model (...) that must be considered close to the truth. I conclude that, since the established techniques for choosing among data models are unequipped to handle these cases, they cannot be regarded as adequate. ‡I presented a previous version of this paper at the 20th Biennial Meeting of the Philosophy of Science Association, Vancouver, November 2006. I am grateful to the audience for constructive discussion. I thank Leiden University students Marjolein Eysink Smeets and Lenneke Schrier for suggesting the cortisol example, and Remko van der Geest for comments on a draft. †To contact the author, please write to: Faculty of Philosophy, University of Leiden, P.O. Box 9515, 2300 RA Leiden, The Netherlands; e-mail: [email protected]. (shrink)

In what follows I will discuss an example of the Duhem-Quine problem in which Pr(H A), Pr(A H), and Pr(OI +H& ?A) (where H is the hypothesis, A the auxiliary assumptions, and O the observational prediction) can be construed objectively; however, only some of those quantities are relevant to the analysis that I provide. The example involves medical diagnosis. The goal is to test the hypothesis that someone has tuberculosis; the auxiliary assumptions describe the er- ror characteristics of the test (...) procedure. Although it can make sense to talk about the objective probability that someone (randomly drawn from a given population) has tuberculosis and it also can make sense to talk about the objective probability that a test procedure has a certain set of error characteristics, neither of these quantities will enter into the analysis. The analysis proceeds entirely via likelihoods, what one needs to consider is just the probability of the observations conditional on four conjunctions of the form (?H & +A).' It is a special feature of the example that all four of these conjunctions are simple statistical hypotheses in the technical sense that each unambigu-ously confers a probability on the observations.'o After describing how the likelihood concept applies to the example concerning medical diagnosis, I will show how similar patterns can arise in the context of a second inferential framework-that of H. Akaike's criterion of model selection;" this time the example will involve phylogenetic inference. (shrink)

ABSTRACTIn his recent book, Rodrik [. Economics rules. Why economics works, when it fails, and how to tell the difference. Oxford University Press] proposes an account of model pluralism according to which multiple models of the same target are acceptable as long as one model is more useful for one purpose and another is more useful for another purpose. How, then, is the right model for the purpose selected? Rodrik roughly outlines a selection procedure, which we (...) formalize to enhance understanding of his account of model pluralism and to advance the critical discussion. (shrink)

This article argues that common intuitions regarding (a) the specialness of ‘use-novel’ data for confirmation and (b) that this specialness implies the ‘no-double-counting rule’, which says that data used in ‘constructing’ (calibrating) a model cannot also play a role in confirming the model’s predictions, are too crude. The intuitions in question are pertinent in all the sciences, but we appeal to a climate science case study to illustrate what is at stake. Our strategy is to analyse the intuitive (...) claims in light of prominent accounts of confirmation of model predictions. We show that on the Bayesian account of confirmation, and also on the standard classical hypothesis-testing account, claims (a) and (b) are not generally true; but for some select cases, it is possible to distinguish data used for calibration from use-novel data, where only the latter confirm. The more specialized classical model-selection methods, on the other hand, uphold a nuanced version of claim (a), but this comes apart from (b), which must be rejected in favour of a more refined account of the relationship between calibration and confirmation. Thus, depending on the framework of confirmation, either the scope or the simplicity of the intuitive position must be revised. (shrink)

When relating genomic data to survival outcomes, there are three main challenges that are the censored survival outcomes, the high-dimensionality of the genomic data, and the non-normality of data. We propose a method to tackle these challenges simultaneously and obtain a robust estimation of detecting significant genes related to survival outcomes based on Accelerated Failure Time model. Specifically, we include a general loss function to the AFT model, adopt model regularization and shrinkage technique, cope with parameters tuning (...) and model selection, and develop an algorithm based on unified Expectation–Maximization approach for easy implementation. Simulation results demonstrate the advantages of the proposed method compared with existing methods when the data has heavy-tailed errors and correlated covariates. Two real case studies on patients are provided to illustrate the application of the proposed method. (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)

In this paper I compare parametric and nonparametric regression models with the help of a simulated data set. Doing so, I have two main objectives. The first one is to differentiate five concepts of simplicity and assess their respective importance. The second one is to show that the scope of the existing philosophical literature on simplicity and model selection is too narrow because it does not take the nonparametric approach into account, S112–S123, 2002; Forster and Sober in The (...) British Journal for the Philosophy of Science 45, 1–35, 1994; Forster, 2001, in Philosophy of Science 74, 588–600, 2007; Hitchcock and Sober in The British Journal for the Philosophy of Science 55, 1–34, 2004; Mikkelson in Philosophy of Science 73, 440–447, 2006; Baker 2013). More precisely, I point out that a measure of simplicity in terms of the number of adjustable parameters is inadequate to characterise nonparametric models and to compare them with parametric models. This allows me to weed out false claims about what makes a model simpler than another. Furthermore, I show that the importance of simplicity in model selection cannot be captured by the notion of parametric simplicity. ‘Simplicity’ is an umbrella term. While parametric simplicity can be ignored, there are other notions of simplicity that need to be taken into consideration when we choose a model. Such notions are not discussed in the previously mentioned literature. The latter therefore portrays an incomplete picture of why simplicity matters when we choose a model. Overall I support a pluralist view according to which we cannot give a general and interesting justification for the importance of simplicity in science. (shrink)

This article argues that common intuitions regarding (a) the specialness of ‘use-novel’ data for confirmation and (b) that this specialness implies the ‘no-double-counting rule’, which says that data used in ‘constructing’ (calibrating) a model cannot also play a role in confirming the model’s predictions, are too crude. The intuitions in question are pertinent in all the sciences, but we appeal to a climate science case study to illustrate what is at stake. Our strategy is to analyse the intuitive (...) claims in light of prominent accounts of confirmation of model predictions. We show that on the Bayesian account of confirmation, and also on the standard classical hypothesis-testing account, claims (a) and (b) are not generally true; but for some select cases, it is possible to distinguish data used for calibration from use-novel data, where only the latter confirm. The more specialized classical model-selection methods, on the other hand, uphold a nuanced version of claim (a), but this comes apart from (b), which must be rejected in favour of a more refined account of the relationship between calibration and confirmation. Thus, depending on the framework of confirmation, either the scope or the simplicity of the intuitive position must be revised. 1 Introduction2 A Climate Case Study3 The Bayesian Method vis-à-vis Intuitions4 Classical Tests vis-à-vis Intuitions5 Classical Model-Selection Methods vis-à-vis Intuitions 5.1 Introducing classical model-selection methods 5.2 Two cases6 Re-examining Our Case Study7 Conclusion. (shrink)

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)

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)

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)

Recent solutions to the curve-fitting problem, described in Forster and Sober ([1995]), trade off the simplicity and fit of hypotheses by defining simplicity as the paucity of adjustable parameters. Scott De Vito ([1997]) charges that these solutions are 'conventional' because he thinks that the number of adjustable parameters may change when the hypotheses are described differently. This he believes is exactly what is illustrated in Goodman's new riddle of induction, otherwise known as the grue problem. However, the 'number of adjustable (...) parameters' is actually a loose way of referring to a quantity that is not language dependent. The quantity arises out of Akaike's theorem in a way that ensures its language invariance. (shrink)

The Akaike Information Criterion can be a valuable tool of scientific inference. This statistic, or any other statistical method for that matter, cannot, however, be the whole of scientific methodology. In this paper some of the limitations of Akaikean statistical methods are discussed. It is argued that the full import of empirical evidence is realized only by adopting a richer ideal of empirical success than predictive accuracy, and that the ability of a theory to turn phenomena into accurate, agreeing measurements (...) of causally relevant parameters contributes to the evidential support of the theory. This is illustrated by Newton's argument from orbital phenomena to the inverse-square law of gravitation. (shrink)

The Akaike Information Criterion can be a valuable tool of scientific inference. This statistic, or any other statistical method for that matter, cannot, however, be the whole of scientific methodology. In this paper some of the limitations of Akaikean statistical methods are discussed. It is argued that the full import of empirical evidence is realized only by adopting a richer ideal of empirical success than predictive accuracy, and that the ability of a theory to turn phenomena into accurate, agreeing measurements (...) of causally relevant parameters contributes to the evidential support of the theory. This is illustrated by Newton's argument from orbital phenomena to the inverse-square law of gravitation. (shrink)

We study Compositional Models based on Dirichlet Regression where, given a (vector) covariate x, one considers the response variable, y, to be a positive vector with a conditional Dirichlet distribution, y | X We introduce a new method for estimating the parameters of the Dirichlet Covariate Model given a linear model on X, and also propose a Bayesian model selection approach. We present some numerical results which suggest that our proposals are more stable and robust than (...) traditional approaches. (shrink)

There has been a lively debate in the philosophy of science over _predictivism_: the thesis that successfully predicting a given body of data provides stronger evidence for a theory than merely accommodating the same body of data. I argue for a very strong version of the thesis using statistical results on the so-called “model selection” problem. This is the problem of finding the optimal model (family of hypotheses) given a body of data. The key idea that I (...) will borrow from the statistical literature is that the level of support a hypothesis, _H,_ receives from a body of data, _D,_ is inversely related to the number of adjustable parameters of the model from which _H_ was constructed. I will argue that when _D_ is not essential to the design of _H_ (i.e., when it is predicted), the model to which _H_ belongs has fewer adjustable parameters than when _D_ is essential to the design of _H_ (when it is accommodated)_._ This, I argue, provides us with an argument for a very strong version of predictivism. (shrink)

We show how the Full Bayesian Significance Test (FBST) can be used as a model selection criterion. The FBST was presented by Pereira and Stern as a coherent Bayesian significance test. Key Words: Bayesian test; Evidence; Global optimization; Information; Model selection; Numerical integration; Posterior density; Precise hypothesis; Regularization. AMS: 62A15; 62F15; 62H15.

The Fully Bayesian Significance Test (FBST) is a coherent Bayesian significance test for sharp hypotheses. This paper proposes the FBST as a model selection tool for general mixture models, and compares its performance with Mclust, a model-based clustering software. The FBST robust performance strongly encourages further developments and investigations.

In herein article an attempt of problem solution connected with anomaly detection in network traffic with the use of statistic models with long or short memory dependence was presented. In order to select the proper type of a model, the parameter describing memory on the basis of the Geweke and Porter-Hudak test was estimated. Bearing in mind that the value of statistic model depends directly on quality of data used for its creation, at the initial stage of the (...) suggested method, outliers were identified and then removed. For the implementation of this task, the criterion using the value of interquartile range was used. The data prepared in this manner were useful for automatic creation of statistic models classes, such as ARFIMA and Holt-Winters. The procedure of calculation of model parameters’ optimal values was carried out as a compromise between the models coherence and the size of error estimation. Then, relations between the estimated network model and its actual parameters were used in order to detect anomalies in the network traffic. Considering the possibility of appearance of significant real traffic network fluctuations, procedure of updating statistic models was suggested. The results obtained in the course of performed experiments proved efficacy and efficiency of the presented solution. (shrink)

Bayesian model selection has frequently been the focus of philosophical inquiry (e.g., Forster, Br J Philos Sci 46:399–424, 1995; Bandyopadhyay and Boik, Philos Sci 66:S390–S402, 1999; Dowe et al., Br J Philos Sci 58:709–754, 2007). This paper argues that Bayesian model selection procedures are very diverse in their inferential target and their justification, and substantiates this claim by means of case studies on three selected procedures: MML, BIC and DIC. Hence, there is no tight link between (...) Bayesian model selection and Bayesian philosophy. Consequently, arguments for or against Bayesian reasoning based on properties of Bayesian model selection procedures should be treated with great caution. (shrink)

Extended redundancy analysis is a statistical method that relates multiple sets of predictors to response variables. In ERA, the conventional approach of model evaluation tends to overestimate the performance of a model since the performance is assessed using the same sample used for model development. To avoid the overly optimistic assessment, we introduce a new model evaluation approach for ERA, which utilizes computer-intensive resampling methods to assess how well a model performs on unseen data. Specifically, (...) we suggest several new model evaluation metrics for ERA that compute a model’s performance on out-of-sample data, i.e., data not used for model development. Although considerable work has been done in machine learning and statistics to examine the utility of cross-validation and bootstrap variants for assessing such out-of-sample predictive performance, to date, no research has been carried out in the context of ERA. We use simulated and real data examples to compare the proposed model evaluation approach with the conventional one. Results show the conventional approach always favor more complex ERA models, thereby failing to prevent the problem of overfitting in model selection. Conversely, the proposed approach can select the true ERA model among many mis-specified models. (shrink)

In psychology, measurement instruments are constructed from scales, which are obtained on the grounds of exploratory and confirmatory factor analysis. Looking at the literature, one can find various recommendations regarding how these techniques should be used during the scale construction process. Some authors suggest to use exploratory factor analysis on the entire data set while others advice to perform an internal cross-validation by randomly splitting the data set in two and then either perform exploratory factor analysis on both parts or (...) exploratory factor analysis on the first part and confirmatory factor analysis on the other. In spite of all these divergent recommendations, there is no consensus on which method yields the best result. In this paper, we analyze this issue in light of the prediction versus accommodation debate and argue that the answer to this question depends on one’s conception of the criteria that should be used to achieve the goals of the scientific enterprise. (shrink)

We’ll adopt a simple framework today. Our assumptions: A model (M) is a family of hypotheses. A hypothesis (H) is a curve plus an associated error term . For simplicity, we’ll assume a common N (0, 1) Gaussian.

The authors have retracted this article [1] because they found a fundamental mistake in the methodology that is not correctable at this time. This mistake is found in the methodology and the derivation of the model with Tukey and Huber’s losses. Because of the error, the findings in the article are not reliable. All authors agree to this retraction.

A phylogeny that allows for lateral gene transfer (LGT) can be thought of as a strictly branching tree (all of whose branches are vertical) to which lateral branches have been added. Given that the goal of phylogenetics is to depict evolutionary history, we should look for the best supported phylogenetic network and not restrict ourselves to considering trees. However, the obvious extensions of popular tree-based methods such as maximum parsimony and maximum likelihood face a serious problem—if we judge networks by (...) fit to data alone, networks that have lateral branches will always fit the data at least as well as any network that restricts itself to vertical branches. This is analogous to the well-studied problem of overfitting data in the curve-fitting problem. Analogous problems often have analogous solutions and we propose to treat network inference as a case of model selection and use the Akaike Information Criterion (AIC). Strictly tree-like networks are more parsimonious than those that postulate lateral as well as vertical branches. This leads to the conclusion that we should not always infer LGT events whenever it would improve our fit-to-data, but should do so only when the improved fit is larger than the penalty for adding extra lateral branches. (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)

Forster and Sober introduced the “sub-family problem” for model selection criteria that recommend balancing goodness-of-fit against simplicity. This problem arises when a maximally simple model is artificially constructed to have excellent fit with the data. We argue that the problem arises because of a violation of the general maxim that balancing goodness-of-fit against simplicity leads to desirable inferences only if one is comparing models for the consideration of which one has a positive reason independently of the current (...) data. (shrink)

In The Biopsychosocial Model of Health and Disease, Derek Bolton and Grant Gillett argue that a defensible updated version of the biopsychosocial model requires a metaphysically adequate account of disease causation that can accommodate biological, psychological, and social factors. This present paper offers a philosophical critique of their account of biopsychosocial causation. I argue that their account relies on claims about the normativity and the semantic content of biological information that are metaphysically contentious. Moreover, I suggest that these (...) claims are unnecessary for a defence of biopsychosocial causation, as the roles of multiple and diverse factors in disease causation can be readily accommodated by a more widely accepted and less metaphysically contentious account of causation. I then raise the more general concern that they are misdiagnosing the problem with the traditional version of the biopsychosocial model. The challenge when developing an explanatorily valuable version of the biopsychosocial model, I argue, is not so much providing an adequate account of biopsychosocial causation, but providing an adequate account of causal selection. Finally, I consider how this problem may be solved to arrive at a more explanatorily valuable and clinically useful version of the biopsychosocial model. (shrink)

This commentary extends Hertwig & Ortmann's analysis by asking how stricter model selection conventions can facilitate the accumulation of information from experimental studies. In many cases researchers are currently motivated to summarize their data with ambiguous and/or multi parameter models. A “generality first” convention can help eliminate this problem.

This chapter examines four solutions to the problem of many models, and finds some fault or limitation with all of them except the last. The first is the naïve empiricist view that best model is the one that best fits the data. The second is based on Popper’s falsificationism. The third approach is to compare models on the basis of some kind of trade off between fit and simplicity. The fourth is the most powerful: Cross validation testing.

We develop an account of laboratory models, which have been central to the group selection controversy. We compare arguments for group selection in nature with Darwin's arguments for natural selection to argue that laboratory models provide important grounds for causal claims about selection. Biologists get information about causes and cause-effect relationships in the laboratory because of the special role their own causal agency plays there. They can also get information about patterns of effects and antecedent conditions (...) in nature. But to argue that some cause is actually responsible in nature, they require an inference from knowledge of causes in the laboratory context and of effects in the natural context. This process, cause detection, forms the core of an analogical argument for group selection. We discuss the differing roles of mathematical and laboratory models in constructing selective explanations at the group level and apply our discussion to the units of selection controversy to distinguish between the related problems of cause determination and evaluation of evidence. Because laboratory models are at the intersection of the two problems, their study is crucial for framing a coherent theory of explanation for evolutionary biology. (shrink)

The key problem in the controversy over group selection is that of defining a criterion of group selection that identifies a distinct causal process that is irreducible to the causal process of individual selection. We aim to clarify this problem and to formulate an adequate model of irreducible group selection. We distinguish two types of group selection models, labeling them type I and type II models. Type I models are invoked to explain differences among (...) groups in their respective rates of production of contained individuals. Type II models are invoked to explain differences among groups in their respective rates of production of distinct new groups. Taking Elliott Sober's model as an exemplar, we argue that although type I models have some biological importance--they force biologists to consider the role of group properties in influencing the fitness of organisms--they fail to identify a distinct group-level causal selection process. Type II models if properly framed, however, do identify a group-level causal selection process that is not reducible to individual selection. We propose such a type II model and apply it to some of the major candidates for group selection. (shrink)

I argue that verbal models should be included in a philosophical account of the scientific practice of modelling. Weisberg (2013) has directly opposed this thesis on the grounds that verbal structures, if they are used in science, only merely describe models. I look at examples from Darwin's On the Origin of Species (1859) of verbally constructed narratives that I claim model the general phenomenon of evolution by natural selection. In each of the cases I look at, a particular (...) scenario is described that involves at least some fictitious elements but represents the salient causal components of natural selection. I pronounce the importance of prioritising observation of scientific practice for the philosophy of modelling and I suggest that there are other likely model types that are excluded from philosophical accounts. (shrink)

Models carry the meaning of science. This puts a tremendous burden on the process of model selection. In general practice, models are selected on the basis of their relative goodness of fit to data penalized by model complexity. However, this may not be the most effective approach for selecting models to answer a specific scientific question because model fit is sensitive to all aspects of a model, not just those relevant to the question. Model (...) Structural Adequacy analysis is proposed as a means to select models based on their ability to answer specific scientific questions given the current understanding of the relevant aspects of the real world. (shrink)

This book publishes 31 of the author's selected papers which have appeared, with one exception, since 1970. The papers cover a wide range of topics in the philosophy of science. Part I is concerned with general methodology, including formal and axiomatic methods in science. Part II is concerned with causality and explanation. The papers extend the author's earlier work on a probabilistic theory of causality. The papers in Part III are concerned with probability and measurement, especially foundational questions about probability. (...) Part IV consists of several papers, including two historical ones, on the foundations of physics, with the main emphasis being on quantum mechanics. Part V, the longest part, is on the foundations of psychology and includes papers mainly on learning and perception. The book is aimed at philosophers of science, scientists concerned with the methodology of the social sciences, and mathematical psychologists interested in theories of learning, perception and measurement. (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)

Cultural variants may spread by being more appealing, more memorable, or less offensive than other cultural variants. Empirical studies suggest that such “emotional selection” is a force to be reckoned with in cultural evolution. We present a research paradigm that is suitable for the study of emotional selection. It guides empirical research by directing attention to the circumstances under which emotions influence the likelihood that an individual will influence another individual to acquire a cultural variant. We present a (...) modeling framework to translate such knowledge into specific and testable predictions of population-level change. A set of already analyzed basic cases can serve as a toolbox. (shrink)

John Burgess is the author of a rich and creative body of work which seeks to defend classical logic and mathematics through counter-criticism of their nominalist, intuitionist, relevantist, and other critics. This selection of his essays, which spans twenty-five years, addresses key topics including nominalism, neo-logicism, intuitionism, modal logic, analyticity, and translation. An introduction sets the essays in context and offers a retrospective appraisal of their aims. The volume will be of interest to a wide range of readers across (...) philosophy of mathematics, logic, and philosophy of language. (shrink)

The epistemic status of Natural Selection has seemed intriguing to biologists and philosophers since the very beginning of the theory to our present times. One prominent contemporary example is Elliott Sober, who claims that NS, and some other theories in biology, and maybe in economics, are peculiar in including explanatory models/conditionals that are a priori in a sense in which explanatory models/conditionals in Classical Mechanics and most other standard theories are not. Sober’s argument focuses on some “would promote” sentences (...) that according to him, play a central role in NS explanations and are both causal and a priori. Lange and Rosenberg criticize Sober arguing that, though there may be some unspecific a priori causal claims, there are not a priori causal claims that specify particular causal factors. Although we basically agree with Lange and Rosenberg’s criticism, we think it remains silent about a second important element in Sober’s dialectics, namely his claim that, contrary to what happens in mechanics, in NS explanatory conditionals are a priori, and that this is so in quite specific explanatory models. In this paper we criticize this second element of Sober’s argument by analyzing what we take to be the four possible interpretations of Sober’s claim, and argue that, terminological preferences aside, the possible senses in which explanatory models in NS can qualify, or include elements that can qualify, as a priori, also apply to CM and other standard, highly unified theories. We conclude that this second claim is unsound, or at least that more needs to be said in order to sustain that NS explanatory models are a priori in a sense in which CM models are not. (shrink)