The presence of an error term does not preclude causal inference in regression: a comment on Krause (2012)

A recent article by Krause (Qual Quant, doi:10.1007/s11135-012-9712-5, Krause (2012)) maintains that: (1) it is untenable to characterize the error term in multiple regression as simply an extraneous random influence on the outcome variable, because any amount of error implies the possibility of one or more omitted, relevant explanatory variables; and (2) the only way to guarantee the prevention of omitted variable bias and thereby justify causal interpretations of estimated coefficients is to construct fully specified models that completely eliminate the error term. The present commentary argues that such an extreme position is impractical and unnecessary, given the availability of specialized techniques for dealing with the primary statistical consequence of omitted variables, namely endogeneity, or the existence of correlations between included explanatory variables and the error term. In particular, the current article discusses the method of instrumental variable estimation, which can resolve the endogeneity problem in causal models where one or more relevant explanatory variables are excluded, thus allowing for accurate estimation of effects. An overview of recent methodological resources and software for conducting instrumental variables estimation is provided, with the aim of helping to place this crucial technique squarely in the statistical toolkit of applied researchers.     

An optimal insurance design problem under Knightian uncertainty

This paper solves an optimal insurance design problem in which both the insurer and the insured are subject to Knightian uncertainty about the loss distribution. The Knightian uncertainty is modeled in a multi-prior g-expectation framework. We obtain an endogenous characterization of the optimal indemnity that extends classical theorems of Arrow (Essays in the Theory of Risk Bearing. Markham, Chicago 1971) and Raviv (Am Econ Rev 69(1):84–96, 1979) in the classical situation. In the presence of Knightian uncertainty, it is shown that the optimal insurance contract is not only contingent on the realized loss but also on another source of uncertainty coming from the ambiguity.     

Index

The Yule distribution is shown to have certain interesting properties in the area of regression analysis. In particular, it is shown that under certain conditions, a random variable Z will have linear regressions on another random variable X and on its observable part Y only when X has a Yule distribution. More generally, the regression on the observed part Y will be constant for a finite number of values of Y, say k, and linear otherwise, only when X has a Yule distribution with its first k frequencies truncated.     

Fuzzy density estimation

A new approach to density estimation with fuzzy random variables (FRV) is developed. In this approach, three methods (histogram, empirical c.d.f., and kernel methods) are extended for density estimation based on α-cuts of FRVs.     

The exact distribution of exogenous variable coefficient estimators

This paper derives the exact probability density function of the instrumental variable (IV) estimator of the exogenous variable coefficient vector in a structural equation containing n + 1 endogenous variables and N degrees of overidentification. The derivations make use of an operator calculus which simplifies the algebra of invariant polynomials with multiple matrix arguments. A leading case of the general distribution that is more amenable to analysis and computation is also presented. Conventional classical assumptions of normally distributed errors and non-random exogenous variables are employed.     

On consistency of the weighted least squares estimators in a semiparametric regression model

This paper is concerned with the semiparametric regression model \(y_i=x_i\beta +g(t_i)+\sigma _ie_i,~~i=1,2,\ldots ,n,\) where \(\sigma _i^2=f(u_i)\), \((x_i,t_i,u_i)\) are known fixed design points, \(\beta \) is an unknown parameter to be estimated, \(g(\cdot )\) and \(f(\cdot )\) are unknown functions, random errors \(e_i\) are widely orthant dependent random variables. The p-th (\(p>0\)) mean consistency and strong consistency for least squares estimators and weighted least squares estimators of \(\beta \) and g under some more mild conditions are investigated. A simulation study is also undertaken to assess the finite sample performance of the results that we established. The results obtained in the paper generalize and improve some corresponding ones of negatively associated random variables.     

Variable selection and functional form uncertainty in cross-country growth regressions

Regression analyses of cross-country economic growth data are complicated by two main forms of model uncertainty: the uncertainty in selecting explanatory variables and the uncertainty in specifying the functional form of the regression function. Most discussions in the literature address these problems independently, yet a joint treatment is essential. We present a new framework that makes such a joint treatment possible, using flexible nonlinear models specified by Gaussian process priors and addressing the variable selection problem by means of Bayesian model averaging. Using this framework, we extend the linear model to allow for parameter heterogeneity of the type suggested by new growth theory, while taking into account the uncertainty in selecting explanatory variables. Controlling for variable selection uncertainty, we confirm the evidence in favor of parameter heterogeneity presented in several earlier studies. However, controlling for functional form uncertainty, we find that the effects of many of the explanatory variables identified in the literature are not robust across countries and variable selections.     

Weak and strong laws of large numbers for arrays of rowwise END random variables and their applications

In the paper, the Marcinkiewicz–Zygmund type moment inequality for extended negatively dependent (END, in short) random variables is established. Under some suitable conditions of uniform integrability, the \(L_r\) convergence, weak law of large numbers and strong law of large numbers for usual normed sums and weighted sums of arrays of rowwise END random variables are investigated by using the Marcinkiewicz–Zygmund type moment inequality. In addition, some applications of the \(L_r\) convergence, weak and strong laws of large numbers to nonparametric regression models based on END errors are provided. The results obtained in the paper generalize or improve some corresponding ones for negatively associated random variables and negatively orthant dependent random variables.     

Information concentration in common value environments

We consider how information concentration affects a seller’s revenue in common value auctions. The common value is a function of $n$ random variables partitioned among $m \le n$ bidders. For each partition, the seller devises an optimal mechanism. We show that whenever the value function allows scalar sufficient statistics for each player’s signals, the mechanism design problem is well-defined. Additionally, whenever a common regularity condition is satisfied, a coarser partition always reduces revenues. In particular, any merger or collusion among bidders reduces revenue.     

Robust games: theory and application to a Cournot duopoly model

In this paper, the robust game model proposed by Aghassi and Bertsimas (Math Program Ser B 107:231–273, 2006) for matrix games is extended to games with a broader class of payoff functions. This is a distribution-free model of incomplete information for finite games where players adopt a robust-optimization approach to contend with payoff uncertainty. They are called robust players and seek the maximum guaranteed payoff given the strategy of the others. Consistently with this decision criterion, a set of strategies is an equilibrium, robust-optimization equilibrium, if each player’s strategy is a best response to the other player’s strategies, under the worst-case scenarios. The aim of the paper is twofold. In the first part, we provide robust-optimization equilibrium’s existence result for a quite general class of games and we prove that it exists a suitable value \(\epsilon \) such that robust-optimization equilibria are a subset of \(\epsilon \)-Nash equilibria of the nominal version, i.e., without uncertainty, of the robust game. This provides a theoretical motivation for the robust approach, as it provides new insight and a rational agent motivation for \(\epsilon \)-Nash equilibrium. In the last part, we propose an application of the theory to a classical Cournot duopoly model which shows significant differences between the robust game and its nominal version.     

Bayesian Estimation of Prediction Error and Variable Selection in Linear Regression

An important statistical application is the problem of determining an appropriate set of input variables for modelling a response variable. In such an application, candidate models are characterized by which input variables are included in the mean structure. A reasonable approach to gauging the propriety of a candidate model is to define a discrepancy function through the prediction error associated with this model. An optimal set of input variables is then determined by searching for the candidate model that minimizes the prediction error. In this paper, we focus on a Bayesian approach to estimating a discrepancy function based on prediction error in linear regression. It is shown how this approach provides an informative method for quantifying model selection uncertainty.     

Robust estimation procedures and visual display techniques for a two-way classification model

An important problem in statistics is to study the effect of one or two factors on a dependent variable. This type of problem can be formulated as a regression problem (by using dummy (0,1) variables to represent the levels of factors) and the standard least squares (LS) analysis is well-known. The least absolute value (LAV) analysis is less well known, but certainly is becoming more widely used, especially in exploratory data analysis.The purpose of this report is to present a didactic treatment of visual display methods useful in exploratory data analysis. These visual display techniques (stem- and- leaf, box- and- whisker, and two-way plots) are presented for both the least squares and the least absolute value analyses of a two-way classification model.     

The concept of weak exchangeability and its applications

A finite sequence of binary random variables is called a weak exchangeable sequence of order m if the sequence consists of m random vectors such that the elements within each random vector are exchangeable in the usual sense and the different random vectors are dependent. The exact and asymptotic joint distributions of the m-dimensional random vector whose elements include the number of successes in each exchangeable sequence are derived. Potential applications of the concept of weak exchangeability are discussed with illustrative examples.     

Random weighting estimation of stable exponent

This paper presents a new random weighting method to estimation of the stable exponent. Assume that $X_1, X_2, \ldots ,X_n$ is a sequence of independent and identically distributed random variables with $\alpha $ -stable distribution G, where $\alpha \in (0,2]$ is the stable exponent. Denote the empirical distribution function of G by $G_n$ and the random weighting estimation of $G_n$ by $H_n$ . An empirical distribution function $\widetilde{F}_n$ with U-statistic structure is defined based on the sum-preserving property of stable random variables. By minimizing the Cramer-von-Mises distance between $H_n$ and ${\widetilde{F}}_n$ , the random weighting estimation of $\alpha $ is constructed in the sense of the minimum distance. The strong consistency and asymptotic normality of the random weighting estimation are also rigorously proved. Experimental results demonstrate that the proposed random weighting method can effectively estimate the stable exponent, resulting in higher estimation accuracy than the Zolotarev, Press, Fan and maximum likelihood methods.     

Strong risk invariance in multiobjective programming

A utility function exhibiting strong risk invariance (SRI) is employed to study a problem in which a decision maker's objectives, y, are linked to his decision variables, x, by the linear system y = Ax, with matrix A random. An application of the model to macroeconomic planning is considered.     

The use of dummy variables to compute predictions,prediction errors,and confidence intervals

This paper presents a method for computing predictions, prediction error variances, and confidence intervals, which can be implemented with any regression program. It demonstrates that a regression estimated for an augmented data set, obtained by (1) combining n sample points with r forecast points, and (2) including r dummy variables (each equalling one only for the corresponding forecast point), will yield r dummy variable coefficients and variances which equal the corresponding prediction errors and prediction error variances. Since most programs lack special routines to calculate these magnitudes, while manual computation is cumbersome, the proposed method is of considerable practical value.     

Extending Rosenberg's idea about conjoint effects

In Rosenberg's work can be found the basis for both a path analytical approach-where relationships between variables are explained—and an R Square measurement—where a variable is explained—for nominal variables. The first approach has been expanded enormously by different authors. In this article we dig into the R Square alike problem. The ideas of predictability and accounted variation—in terms of percentage changed over time—and a simple statistic like percentages are on the basis of the measurements that we present, with a similar aim as the R square, but applied to contingency tables. For cross-sectional data this measurement is based on the percentage of cases in a category of the dependent variable for the particular combination of categories of the predictor ones making this figure as big as possible. We call this measurement the Coefficient of Predictability, P Square, and its value can be interpreted as the percentage of unpredictability which is accounted for by the variables in the model. With change data we try to explain why the percentage of cases in a category of a variable changes over time. This means that we have to look at the relationship between time and this variable, controlling for another set of variables. The variation is explained when the relationship gets down to zero, and the closer to this value, the better the explanatory power of the variables in our model. We call D Square to this measurement of accounted variation.     

A new measure of association between random variables

We propose a new measure of association between two continuous random variables X and Y based on the covariance between X and the log-odds rate associated to Y. The proposed index of correlation lies in the range [\(-1\), 1]. We show that the extremes of the range, i.e., \(-1\) and 1, are attainable by the Fr\(\acute{\mathrm{e}}\)chet bivariate minimal and maximal distributions, respectively. It is also shown that if X and Y have bivariate normal distribution, the resulting measure of correlation equals the Pearson correlation coefficient \(\rho \). Some interpretations and relationships to other variability measures are presented. Among others, it is shown that for non-negative random variables the proposed association measure can be represented in terms of the mean residual and mean inactivity functions. Some illustrative examples are also provided.     

Some characterizations of distributions by regression models for ordinal response data

Summary Let a random variableX be classified intok classes. By doing so, a new random variable is obtained, measured on ordinal scale. If this variable is a response variable in certain regression models for ordinal response data, the distribution ofX is characterized by the models. In this paper, characterizations of the distribution ofX by the proportional odds model and the proportional hazards model are given.     

The LeChatelier principle for changes in risk

In this paper, we analyze the comparative statics of changes in risk in the context of problems with multiple decision variables. We demonstrate, in particular, that the Samuelson–LeChatelier principle extends naturally to the comparative statics of changes in risk: in the presence of positive feedbacks between the decision variables, the unrestricted response to an increase in risk is in the same direction and stronger in magnitude than the restricted response (i.e. the choice when other decision variables are fixed). We define the concepts of Nth-degree risk complements and Nth-degree risk substitutes and we show that it is in any one of these two cases (and only in these cases) that we will observe positive feedbacks between the decision variables. We also analyze the extent to which the same principle can be applied to strategic settings under uncertainty.