Inference in Censored Models with Endogenous Regressors

This paper analyzes the linear regression model y = xβ+ε with a conditional median assumption med (ε| z) = 0, where z is a vector of exogenous instrument random variables. We study inference on the parameter β when y is censored and x is endogenous. We treat the censored model as a model with interval observation on an outcome, thus obtaining an incomplete model with inequality restrictions on conditional median regressions. We analyze the identified features of the model and provide sufficient conditions for point identification of the parameter β. We use a minimum distance estimator to consistently estimate the identified features of the model. We show that under point identification conditions and additional regularity conditions, the estimator based on inequality restrictions is normal and we derive its asymptotic variance. One can use our setup to treat the identification and estimation of endogenous linear median regression models with no censoring. A Monte Carlo analysis illustrates our estimator in the censored and the uncensored case.     

Estimating Health Risks from Natural Hazards Using Risk Assessment and Epidemiology

Risk assessment is the process of estimating the likelihood that an adverse effect may result from exposure to a specific health hazard. The process traditionally involves hazard identification, dose-response assessment, exposure assessment, and risk characterization to answer “How many excess cases of disease A will occur in a population of size B due to exposure to agent C at dose level D?” For natural hazards, however, we modify the risk assessment paradigm to answer “How many excess cases of outcome Y will occur in a population of size B due to natural hazard event E of severity D?” Using a modified version involving hazard identification, risk factor characterization, exposure characterization, and risk characterization, we demonstrate that epidemiologic modeling and measures of risk can quantify the risks from natural hazard events. We further extend the paradigm to address mitigation, the equivalent of risk management, to answer “What is the risk for outcome Y in the presence of prevention intervention X relative to the risk for Y in the absence of X?” We use the preventable fraction to estimate the efficacy of mitigation, or reduction in adverse health outcomes as a result of a prevention strategy under ideal circumstances, and further estimate the effectiveness of mitigation, or reduction in adverse health outcomes under typical community-based settings. By relating socioeconomic costs of mitigation to measures of risk, we illustrate that prevention effectiveness is useful for developing cost-effective risk management options.     

Near optimal solutions for maximum quasi-bicliques

The maximum quasi-biclique problem has been proposed for finding interacting protein group pairs from large protein-protein interaction (PPI) networks. The problem is defined as follows: The Maximum Quasi-biclique Problem: Given a bipartite graph G=(X∪Y,E) and a number 0<δ≤0.5, find a subset X _opt of X and a subset Y _opt of Y such that any vertex x∈X _opt is incident to at least (1?δ)|Y _opt | vertices in Y _opt , any vertex y∈Y _opt is incident to at least (1?δ)|X _opt | vertices in X _opt and |X _opt |+|Y _opt | is maximized. The problem was proved to be NP-hard. We design a polynomial time approximation scheme to give a quasi-biclique X′?X and Y′?Y with |X′|+|Y′|≥(1?ε)(|X _opt |+|Y _opt |) such that any vertex x∈X′ is incident to at least (1?δ?ε)|Y′| vertices in Y′ and any vertex y∈Y′ is incident to at least (1?δ?ε)|X′| vertices in X′ for any ε>0, where X _opt and Y _opt form the optimal solution.     

Panel Data Models With Interactive Fixed Effects

This paper considers large N and large T panel data models with unobservable multiple interactive effects, which are correlated with the regressors. In earnings studies, for example, workers' motivation, persistence, and diligence combined to influence the earnings in addition to the usual argument of innate ability. In macroeconomics, interactive effects represent unobservable common shocks and their heterogeneous impacts on cross sections. We consider identification, consistency, and the limiting distribution of the interactive‐effects estimator. Under both large N and large T, the estimator is shown to be consistent, which is valid in the presence of correlations and heteroskedasticities of unknown form in both dimensions. We also derive the constrained estimator and its limiting distribution, imposing additivity coupled with interactive effects. The problem of testing additive versus interactive effects is also studied. In addition, we consider identification and estimation of models in the presence of a grand mean, time‐invariant regressors, and common regressors. Given identification, the rate of convergence and limiting results continue to hold.     

On the generalized constrained longest common subsequence problems

We investigate four variants of the longest common subsequence problem. Given two sequences X, Y and a constrained pattern P of lengths m, n, and ρ, respectively, the generalized constrained longest common subsequence (GC-LCS) problems are to find a longest common subsequence of X and Y including (or excluding) P as a subsequence (or substring). We propose new dynamic programming algorithms for solving the GC-LCS problems in O(mn ρ) time. We also consider the case where the number of constrained patterns is arbitrary.     

Testing for Causal Effects in a Generalized Regression Model With Endogenous Regressors

A unifying framework to test for causal effects in nonlinear models is proposed. We consider a generalized linear‐index regression model with endogenous regressors and no parametric assumptions on the error disturbances. To test the significance of the effect of an endogenous regressor, we propose a statistic that is a kernel‐weighted version of the rank correlation statistic (tau) of Kendall (1938). The semiparametric model encompasses previous cases considered in the literature (continuous endogenous regressors (Blundell and Powell (2003)) and a single binary endogenous regressor (Vytlacil and Yildiz (2007))), but the testing approach is the first to allow for (i) multiple discrete endogenous regressors, (ii) endogenous regressors that are neither discrete nor continuous (e.g., a censored variable), and (iii) an arbitrary “mix” of endogenous regressors (e.g., one binary regressor and one continuous regressor).     

Identification and Estimation of Average Partial Effects in “Irregular” Correlated Random Coefficient Panel Data Models

In this paper we study identification and estimation of a correlated random coefficients (CRC) panel data model. The outcome of interest varies linearly with a vector of endogenous regressors. The coefficients on these regressors are heterogenous across units and may covary with them. We consider the average partial effect (APE) of a small change in the regressor vector on the outcome (cf. Chamberlain (1984), Wooldridge (2005a)). Chamberlain (1992) calculated the semiparametric efficiency bound for the APE in our model and proposed a √N‐consistent estimator. Nonsingularity of the APE's information bound, and hence the appropriateness of Chamberlain's (1992) estimator, requires (i) the time dimension of the panel (T) to strictly exceed the number of random coefficients (p) and (ii) strong conditions on the time series properties of the regressor vector. We demonstrate irregular identification of the APE when T = p and for more persistent regressor processes. Our approach exploits the different identifying content of the subpopulations of stayers—or units whose regressor values change little across periods—and movers—or units whose regressor values change substantially across periods. We propose a feasible estimator based on our identification result and characterize its large sample properties. While irregularity precludes our estimator from attaining parametric rates of convergence, its limiting distribution is normal and inference is straightforward to conduct. Standard software may be used to compute point estimates and standard errors. We use our methods to estimate the average elasticity of calorie consumption with respect to total outlay for a sample of poor Nicaraguan households.     

Confidence Intervals for Diffusion Index Forecasts and Inference for Factor‐Augmented Regressions

We consider the situation when there is a large number of series, N, each with T observations, and each series has some predictive ability for some variable of interest. A methodology of growing interest is first to estimate common factors from the panel of data by the method of principal components and then to augment an otherwise standard regression with the estimated factors. In this paper, we show that the least squares estimates obtained from these factor‐augmented regressions are consistent and asymptotically normal if . The conditional mean predicted by the estimated factors is consistent and asymptotically normal. Except when T/N goes to zero, inference should take into account the effect of “estimated regressors” on the estimated conditional mean. We present analytical formulas for prediction intervals that are valid regardless of the magnitude of N/T and that can also be used when the factors are nonstationary.     

Estimation and Inference in Large Heterogeneous Panels with a Multifactor Error Structure

This paper presents a new approach to estimation and inference in panel data models with a general multifactor error structure. The unobserved factors and the individual‐specific errors are allowed to follow arbitrary stationary processes, and the number of unobserved factors need not be estimated. The basic idea is to filter the individual‐specific regressors by means of cross‐section averages such that asymptotically as the cross‐section dimension (N) tends to infinity, the differential effects of unobserved common factors are eliminated. The estimation procedure has the advantage that it can be computed by least squares applied to auxiliary regressions where the observed regressors are augmented with cross‐sectional averages of the dependent variable and the individual‐specific regressors. A number of estimators (referred to as common correlated effects (CCE) estimators) are proposed and their asymptotic distributions are derived. The small sample properties of mean group and pooled CCE estimators are investigated by Monte Carlo experiments, showing that the CCE estimators have satisfactory small sample properties even under a substantial degree of heterogeneity and dynamics, and for relatively small values of N and T.     

Minimizing the maximum bump cost in linear extensions of a poset

A linear extension of a poset P=(X,?) is a permutation x ₁,x ₂,…,x _|X| of X such that i<j whenever x _i ?x _j . For a given poset P=(X,?) and a cost function c(x,y) defined on X×X, we want to find a linear extension of P such that maximum cost is as small as possible. For the general case, it is NP-complete. In this paper we consider the linear extension problem with the assumption that c(x,y)=0 whenever x and y are incomparable. First, we prove the discussed problem is polynomially solvable for a special poset. And then, we present a polynomial algorithm to obtain an approximate solution.     

Comprehensive Education for the Unemployed — Evaluating the Effects on Unemployment of the Adult Education Initiative in Sweden

Abstract. This paper evaluates the effects on unemployment in Sweden of the Adult Education Initiative (AEI) which during its run from 1997 to 2002 offered adult education to the unemployed at compulsory or upper secondary level. The AEI is compared with the vocational part of Labor Market Training (LMT) using unemployment incidence and unemployment duration as outcome variables, both measured immediately after completion of the programs. For unemployment incidence, selection on unobservables is taken into account by using a bivariate probit model. The analysis of unemployment duration considers both selection bias and censored observations. The results indicate lower incidence following participation in the AEI, but also — significant at the 10 per cent level — longer duration.     

Efficient Semiparametric Estimation via Moment Restrictions

Conditional moment restrictions can be combined through GMM estimation to construct more efficient semiparametric estimators. This paper is about attainable efficiency for such estimators. We define and use a moment tangent set, the directions of departure from the truth allowed by the moments, to characterize when the semiparametric efficiency bound can be attained. The efficiency condition is that the moment tangent set equals the model tangent set. We apply these results to transformed, censored, and truncated regression models, e.g., finding that the conditional moment restrictions from Powell's (1986) censored regression quantile estimators can be combined to approximate efficiency when the disturbance is independent of regressors.     

Geometry of Semidefinite Max-Cut Relaxations via Matrix Ranks

Semidefinite programming (SDP) relaxations are proving to be a powerful tool for finding tight bounds for hard discrete optimization problems. This is especially true for one of the easier NP-hard problems, the Max-Cut problem (MC). The well-known SDP relaxation for Max-Cut, here denoted SDP1, can be derived by a first lifting into matrix space and has been shown to be excellent both in theory and in practice.Recently the present authors have derived a new relaxation using a second lifting. This new relaxation, denoted SDP2, is strictly tighter than the relaxation obtained by adding all the triangle inequalities to the well-known relaxation. In this paper we present new results that further describe the remarkable tightness of this new relaxation. Let denote the feasible set of SDP2 for the complete graph with n nodes, let F _n denote the appropriately defined projection of into , the space of real symmetric n × n matrices, and let C _n denote the cut polytope in . Further let be the matrix variable of the SDP2 relaxation and X F _n be its projection. Then for the complete graph on 3 nodes, F ₃ = C ₃ holds. We prove that the rank of the matrix variable of SDP2 completely characterizes the dimension of the face of the cut polytope in which the corresponding matrix X lies. This shows explicitly the connection between the rank of the variable Y of the second lifting and the possible locations of the projected matrix X within C ₃. The results we prove for n = 3 cast further light on how SDP2 captures all the structure of C ₃, and furthermore they are stepping stones for studying the general case n 4. For this case, we show that the characterization of the vertices of the cut polytope via rank Y = 1 extends to all n 4. More interestingly, we show that the characterization of the one-dimensional faces via rank Y = 2 also holds for n 4. Furthermore, we prove that if rank Y = 2 for n 3, then a simple algorithm exhibits the two rank-one matrices (corresponding to cuts) which are the vertices of the one-dimensional face of the cut polytope where X lies.     

End‐of‐Sample Instability Tests

This paper considers tests for structural instability of short duration, such as at the end of the sample. The key feature of the testing problem is that the number, m, of observations in the period of potential change is relatively small—possibly as small as one. The well‐known F test of Chow (1960) for this problem only applies in a linear regression model with normally distributed iid errors and strictly exogenous regressors, even when the total number of observations, n+m, is large. We generalize the F test to cover regression models with much more general error processes, regressors that are not strictly exogenous, and estimation by instrumental variables as well as least squares. In addition, we extend the F test to nonlinear models estimated by generalized method of moments and maximum likelihood. Asymptotic critical values that are valid as n→∞ with m fixed are provided using a subsampling‐like method. The results apply quite generally to processes that are strictly stationary and ergodic under the null hypothesis of no structural instability.     

Selective support vector machines

In this study we introduce a generalized support vector classification problem: Let X _i , i=1,…,n be mutually exclusive sets of pattern vectors such that all pattern vectors x _i,k , k=1,…,|X _i | have the same class label y _i . Select only one pattern vector $x_{i,k^{*}}In this study we introduce a generalized support vector classification problem: Let X _i , i=1,…,n be mutually exclusive sets of pattern vectors such that all pattern vectors x _i,k , k=1,…,|X _i | have the same class label y _i . Select only one pattern vector from each set X _i such that the margin between the set of selected positive and negative pattern vectors are maximized. This problem is formulated as a quadratic mixed 0-1 programming problem, which is a generalization of the standard support vector classifiers. The quadratic mixed 0-1 formulation is shown to be -hard. An alternative approach is proposed with the free slack concept. Primal and dual formulations are introduced for linear and nonlinear classification. These formulations provide flexibility to the separating hyperplane to identify the pattern vectors with large margin. Iterative elimination and direct selection methods are developed to select such pattern vectors using the alternative formulations. These methods are compared with a na?ve method on simulated data. The iterative elimination method is also applied to neural data from a visuomotor categorical discrimination task to classify highly cognitive brain activities.     

Identification of Marginal Effects in Nonseparable Models Without Monotonicity

Nonseparable models do not impose any type of additivity between the unobserved part and the observable regressors, and are therefore ideal for many economic applications. To identify these models using the entire joint distribution of the data as summarized in regression quantiles, monotonicity in unobservables has frequently been assumed. This paper establishes that in the absence of monotonicity, the quantiles identify local average structural derivatives of nonseparable models.     

Distance graphs on R n with 1-norm

Suppose S is a subset of a metric space X with metric d. For each subset D⊆{d(x,y):x,y∈S,x≠y}, the distance graph G(S,D) is the graph with vertex set S and edge set E(S,D)={xy:x,y∈S,d(x,y)∈D}. The current paper studies distance graphs on the n-space R ₁ ⁿ with 1-norm. In particular, most attention is paid to the subset Z ₁ ⁿ of all lattice points of R ₁ ⁿ . The results obtained include the degrees of vertices, components, and chromatic numbers of these graphs. Dedicated to Professor Frank K. Hwang on the occasion of his 65th birthday. Supported in part by the National Science Council under grant NSC-94-2115-M-002-015. Taida Institue for Mathematical Sciences, National Taiwan University, Taipei 10617, Taiwan. National Center for Theoretical Sciences, Taipei Office.     

The Effectiveness of Graphical Presentation for Information Extraction: A Cumulative Experimental Approach*

Three laboratory experiments were conducted to assess the relative strengths and weaknesses of bar, symbol, and line graphs for performing a variety of elementary information extraction tasks using two dependent variables, time and accuracy. The findings indicate that the degree of support provided by a particular graph format for a particular data extraction task depends on the matching of format and task in terms of their anchoring characteristics. Anchoring, in this context, refers to the phenomenon that specific and diverse parts of a graph are segmented by the reader to act as salient and relevant cues, or anchors, when different classes of information are to be extracted from the graph. A data extraction task has high x-value (y-value) anchoring if the x-axis (y-axis) component is represented in the question as either a given value or an unknown value. Conversely, a task has low x-value (y-value) anchoring if the x-axis (y-axis) component is not represented in the question as either a given value or as an unknown value. Data extraction accuracy was not significantly affected by presentation format. Bars provided the best time performance for data extraction tasks having high anchoring on both axes but were not appropriate for tasks having low anchoring on either the y axis or both the x and y axes. Line graphs tended to be worse in terms of time performance for tasks having high anchoring on both axes although they were as fast or better than other representations for tasks having low anchoring on both axes. Symbol plots appeared to possess anchoring characteristics associated with both bars and line graphs. Symbols (as with bars) tended to produce a time performance superior to that of line graphs for tasks having high anchoring on both axes; and (as with line graphs) symbols allowed faster results than bar graphs for tasks having low anchoring on either the y axis or both the x and y axes.     

Shift Restrictions and Semiparametric Estimation in Ordered Response Models

We develop a √n‐consistent and asymptotically normal estimator of the parameters (regression coefficients and threshold points) of a semiparametric ordered response model under the assumption of independence of errors and regressors. The independence assumption implies shift restrictions allowing identification of threshold points up to location and scale. The estimator is useful in various applications, particularly in new product demand forecasting from survey data subject to systematic misreporting. We apply the estimator to assess exaggeration bias in survey data on demand for a new telecommunications service.