Nonparametric Frontier Estimation: The Smooth Local Maximum Estimator and The Smooth FDH Estimator
|
|
- Austen Turner
- 6 years ago
- Views:
Transcription
1 Nonparametric Frontier Estimation: The Smooth Local Maximum Estimator and The Smooth FDH Estimator Hudson da S. Torrent 1 July, 2013 Abstract. In this paper we propose two estimators for deterministic production frontier models named as Max- Smooth and Smooth-FDH. The referred estimators have desirable properties: (i) they are smooth; (ii) they are not inherent biased as FDH and DEA estimators, (iii) they are simple to implement and (iv) they pursue less variance than estimators based on conditional moments. Max-Smooth estimator is constructed in three steps. The first step is a novel local maximum estimator. The second step is responsible to smooth out the variance presented in the first step. Then a third step is proposed to correct the position of the estimated frontier. Smooth-FDH is similar to Max-Smooth but with the first stage replaced by the FDH estimator. A cross-validation procedure for bandwidth selection is presented. We compare the relative performance of the proposed estimators with DEA and FDH in a simulation study. The results are very favorable to the proposed estimators even in a simulation setup that is very suitable for DEA estimator. Keywords and phrases. nonparametric frontier models; local smoothing; local linear regression. JEL Classifications. C14, C21 Area. Econometria 1 Department of Statistics and PPGE, Federal University of Rio Grande do Sul, Porto Alegre - Brazil, hudson.torrent@ufrgs.br.
2 1 Introduction Estimation of production frontiers and therefore efficiency (and inefficiency) of production processes has been the subject of a vast and growing literature since Farrell (1957). The problem can be stated as follows. Let x R p + be a set of inputs used to produce a set of outputs y R q +. So, there is a technological or production set defined as Ψ = {(x, y) R p+q + x can produce y}. A production frontier associated with Ψ is defined as ρ(x) = sup{y R q + (y, x) Ψ} for all x R p +. Thus, for given (x 0, y 0 ) Ψ, efficiency is measured by the distance between y 0 and ρ(x 0 ). In this problem we intend to estimate from a given random sample χ = {(x i, y i ), i = 1,..., n} an associated production frontier ρ( ) and efficiency measures for the observed production units 1. To solve these problems we can find in the literature two traditional nonparametric estimation procedures, Free Disposal Hull (FDH) estimator that was introduced by Deprins et al. (1984) and Data Envelopment Analysis (DEA), represented by Charnes et al. (1978). The idea is to estimate a production set from an observed random sample without being necessary to assume any restrictive parametric structure either on the production frontier ρ( ) or on the joint density of (X i, Y i ). Many works apply these methodologies. Gijbels et al. (1999) and Park et al. (2000) have obtained asymptotic distributions for DEA and FDH estimators, respectively. However, these estimators have some characteristics that may be undesirable. Both estimators consist of obtaining the smallest set that envelops all data with the restriction of free disposability (FDH and DEA) and convexity (DEA). Therefore the estimated set never exceeds the true frontier; hence the frontier estimators are inherently downwards biased. Moreover, FDH produces a discontinuous function that envelops the data and DEA produces a piecewise linear function. Martins-Filho and Yao (2007) propose a deterministic production frontier model and a nonparametric production frontier estimator in three stages. First step is estimating a conditional mean using local linear Kernel estimation. The second step follows Fan and Yao (1998), i.e., a local linear Kernel method is used to estimate the conditional variance function. The third and final step is an original estimator that is relative to their proposed production frontier model. The estimation is based on estimating the conditional variance (and the square root of the estimatives) in order to get the shape of the frontier. They derive the asymptotic normality and consistency of both production frontier and efficiency estimators under reasonable assumptions in the nonparametric context. This estimator shares the flexible nonparametric structure, moreover it has some extra desirable properties if compared to FDH and DEA estimators: i) the frontier estimator is a smooth function of input usage (not discontinuous neither piecewise linear) and ii) although the estimator envelops the data it is not inherently biased as FDH and DEA estimators. However, an undesirable result may be emerging in the second step of that estimator since the estimation procedure allows for a negative estimate of the variance. Furthermore, the referred estimator presents a pronounced variance since its based on conditional moments instead of conditional maximum. Therefore we propose in this paper an estimator that has advantages over the estimators cited above. Regarding 1 We consider here only the deterministic approach for efficiency problem. Within this approach all observations are supposed to lie inside the technological set. 1
3 DEA and FDH, our estimator is interesting since it incorporates the smoothness of nonparametric kernel methodology. Likewise, it has advantages over the estimator proposed by Martins-Filho and Yao (2007) since it is based on conditional maximum, which results in a lower variance of the proposed estimator. From here on we denoted our estimator by Max- Smooth. This estimator is characterized by three steps as follows. First we estimate the maximum output conditional on a given input value via a simple non-smooth nonparametric procedure. In the second stage we smooth out the variance of the first stage by making use of nonparametric kernel regression. Finally we correct the position of the estimated frontier, making it above all data points. We also consider an alternative version of Max-Smooth, denoted by FDH-Smooth. This version is characterized by substituting the first stage by the FDH estimator. This approach has the convenience of that just one bandwidth is needed to implement the estimator and it takes advantage of free disposability. Clearly, it is less flexible than Max-Smooth, which results in worse performance in some situations as we shall see in section 4. This paper is composed as follows. In the second section we present the model and the estimation procedure, giving details about our proposed estimators. In Section 3 we propose a cross-validation type procedure for bandwidth selection. In Section 4 the simulation study is presented in detail and we establish comparison among the proposed estimators and FDH and DEA. Finally, in Section 5 conclusions and final comments are stated. 2 Stochastic Model and Estimation Procedure In this section we present the stochastic model and the estimation procedure for that model. The problem may be viewed considering a firm that makes only one product from k inputs, that is, (x, y) R p + R +, where x describes p inputs used for production and y describes the output (one-output case) of a production unit. The production set is defined as previously. In a unique product case we have the following: Ψ = {(x, y) R p+1 + x can produce y} The production function or frontier associated with Ψ is ρ(x) = sup{y R q + (y, x) Ψ} for all x R p +. In practice Ψ and its frontier are unknown, so our prior interest is estimating this frontier from a set of observed firms, i.e., given a random sample of production units {(X i, Y i )} n i=1 that share a technology Ψ, obtaining estimates of ρ( ). By extension we are interested in constructing efficiency ranks and relative performance of production units. To see this, let (x 0, y 0 ) Ψ characterize the performance of a production unit and define 0 R 0 unit s (inverse) Farrell output efficiency measure. 2 From estimates of ρ we can obtain estimates of R 0. y0 ρ(x 0) 1 to be this We propose to estimate the frontier using nonparametric methods as follows. First we estimate the maximum output conditional on a given input value via a simple non-smooth nonparametric procedure. Denote the estimated 2 Note that if the production level y 0 associated with x 0 lies on the frontier function we have y 0 = ρ(x 0 ). The production process is efficient and R 0 = 1. 2
4 maximum output value for unit i as Yi max. The proposed estimator is ( Y max i = max 1 j n Y j I [ 1,1] ( Xj X i h 1 )), i = 1,, n; (1) where I is an indicator function and h 1 > 0 is a bandwidth. This first stage may be viewed as estimating those values of output that better represent more efficient units for a given level of input. The sequence Y max i follows. is then viewed as where E(u i X i ) 0. Now, let µ u := E(u i X i ). Therefore, Y max i = ρ(x i ) + u i, i = 1,, n; Y max i = m(x i ) + ɛ i, i = 1,, n; (2) where E(ɛ i X i ) E(u i µ u X i ) = 0, V ar(ɛ X i ) = σ 2 (X i ) and m(x i ) = ρ(x i ) + µ u. Equation 2 is suitable for nonparametric regression. We use the local linear Kernel estimator of Fan (1992) with regressand Y max i X i. That is, for any x R p + we obtain ˆm(x) ˆα where (ˆα, ˆβ) = arg min α,β n i=1 and regressors (Y max i α β(x i x)) 2 K h2 (X i x) (3) K( ) : R p R is a symmetric density function, K h (u) = (1/h)K(u/h) and h 2 > 0 is a bandwidth. Since our main interest lies on estimating ρ( ), we propose to estimate ˆµ u by Hence, we have for any x R p +, ˆµ u = max 1 i n (Y i ˆm(X i )), (4) ˆρ(x) = ˆm(x) ˆµ u. (5) Remark 1. The main purpose of the first step (eq. (1)) is to get the shape of frontier even though in a wrong position and without smoothness. Note that the proposed estimator is very flexible in the sense of not imposing any particular assumption on frontier s shape. From here on we call this estimator as Max-Smooth. Nevertheless, imposing some structure or restriction on the first step may be interesting if the restriction is in accordance with the DGP. In this regard, free disposability seems to be a reasonable assumption to be assumed about the technology. Therefore, we also consider an estimator that incorporates the free disposability in the first step. In order to do so we propose to use FDH estimator in place of eq. (1), keeping the other steps unchanged. From here on we call this estimator as Smooth-FDH. Clearly the Smooth-FDH estimator has the advantage of requiring just one bandwidth. Furthermore, for those DGP s characterized by strict monotonicity that estimator is more likely to show better performance than Max-Smooth estimator. On the other hand, we conjecture that the referred flexibility in the first step is valuable in some situations. For instance, when the frontier is not too monotone. In Figures (1) and (2) we give some examples of how the estimators described in remark 1 look like in general. The production frontiers used in the examples are considered in a simulation study described in Section 4. 3
5 Figura 1: Frontier I - Ilustration regarding Max-Smooth and Smooth-FDH estimators Max-Smooth - Smooth-FDH - y True 1st Stage 2nd Stage Max-Smooth y True FDH 2nd Stage Smooth-FDH x x Remark 2. Alternatively, it is possible to estimate the frontier in just two steps, eliminating the positioning step. This would be accomplished by different choices for the bandwidths. Notably, h 1 would be bigger (than that for the 3-stage procedure) in order to get at the same stage shape and position of the frontier. Two possible drawbacks in this approach are (i) there is no guarantee that all observed points lie bellow the estimated frontier; (ii) it seems that estimating the frontier in just two steps result in losing precision, at least in finite sample, as we shall see in a simulation study in section 4. On the other hand, we conjecture that the 2-stage estimator may be more robust to extreme values, since it is not restricted to be above all data points. We shall consider outlier scenario in a future version of the paper. Figura 2: Frontier II - Ilustration regarding Max-Smooth and Smooth-FDH estimators Max-Smooth - Smooth-FDH - y True 1st Stage 2nd Stage Max-Smooth y True FDH 2nd Stage Smooth-FDH x x 4
6 3 Bandwidth Selection We need to estimate h 1 and h 2 in order to implement our estimator. In this section we describe a simple cross-validation type procedure for selecting those bandwidths. We select h 1 and h 2 over a grid of points. The algorithm for bandwidth selection may be described as follows. 1. Select a candidate for h 1 and estimate the first stage (eq. (2)). 2. For all candidates for h 2 estimate the second stage (eq. (3)), but for each X i excluding the correspondent Y max i (and all Y max with the same value if that is the case). 3. Estimate the third stage (eq. (4)) in order to obtain the cross-validation estimated frontier, denoted here by ˆρ CV ( ). 4. Evaluate the following function n CV (h 1, h 2 ) = (Y i ˆρ CV (X i )) 2 (6) i=1 5. Repeat this process for all candidates for h 1. Pick h 1 and h 2 that minimize eq. (6). Note that the exclusion of Y max in 2 is important to avoid selecting a pair of bandwidths that in practice would interpolate the points. Furthermore, the validity of eq. (6) lies on the fact that all sample points are below the estimated frontier by construction of our estimator. 4 Simulations In this section we attempt to highlight the finite sample properties of our estimator. We consider the following DGP: Y i = σ(x i) σ R R i, with p = 1, where, X i are pseudo random variables with uniform distribution on [a, b] where a, b are specified in eq. (7) bellow. R i = exp( Z i ), where Z i are pseudo random variables from an exponential distribution with parameter λ. We consider λ = 3. This parameter for the exponential distribution results in mean efficiency of We consider two specifications for ρ( ): Frontier I: ρ 1 (x) = (x) with x [4, 25], and (7) Frontier II: ρ 2 (x) = 3(x 1.5) x with x [1, 2]. In order to evaluate performance of the estimators we consider two measures of error. One of them is denoted by MSE and is defined as n ( 2 MSE(ˆρ) j = ˆρ(X i ) ρ(x i )), j = 1,, n r, (8) i=1 5
7 where ˆρ( ) is a given estimator and n r is the number of repetitions. We present boxplots concerning this measure for each estimator in each case considered. Moreover we present boxplots of the error committed by an estimator around the true value of the frontier evaluated in a given point. In more detail, we define Err(ˆρ) j = ˆρ(x k ) ρ(x k ), k = 1,, K; j = 1,, n r. (9) We consider K = 3 that correspond to the 0.1, 0.5 and 0.9 quantiles of X. That is, regarding frontier I we have x 1 = 6.1, x 2 = 14.5 and x 3 = For frontier II we have x 1 = 1.1, x 2 = 1.5 and x 3 = 1.9. We consider three sample sizes, 200, 400; and n r = 1000 for each one of the experiments. 4.1 Simulation I In the first set of simulations we investigate the questions pointed out in Remark 1 and Remark 2 in section 2. We consider three estimators. The 3-stage Max-smooth estimator; the 3-stage Smooth-FDH estimator; and a 2-stage estimator, denoted by MaxS-2S. In order to focus on the performance of the estimators we consider an oracle bandwidth selection for the three estimators considered in this subsection. The oracle bandwidths are those that minimize the MSE for each sample, letting the true frontier to be known. This study attempts to highlight to properties of the estimators without concerning about the bandwidth selection and therefore about the error incurred by bandwidth selection. The results are presented and analyzed in subsection 4.2 bellow. 4.2 Simulation II In the second set of simulations we attempt to shed some light on the performance of the proposed estimators in practice. That is, we now consider the performance of the estimators using a data-driven procedure to select the bandwidths. In particular, we are interested in investigating wether the presence of two bandwidths could hidden the applicability of Max-Smooth estimator. We also include the traditional FDH estimator for both frontiers and the traditional DEA estimator for Frontier I. The results are presented and analyzed in the next subsection. 4.3 Analysis of the results Now we present and comment the results from the simulations described above Simulation I Frontier I: (Figures (3) - (10)). We see that the performance of all estimators gets better as n increases. The Smooth- FDH presents the best performance in all cases considered while MaxS-2S presents the worst one. It seems that a 3-stage estimator is preferable than a 2-stage estimator. Regarding Max-Smooth and Smooth-FDH, it seems that the second takes advantage of the free disposability in this case, since the DGP pursue that characteristic. Frontier II: In this scenario the best performance is presented by Max-Smooth estimator followed by MAxS-2S. This seems to be explained by a relatively poor performance of Smooth-FDH especially in the flat portion of the frontier, as we see on Figures (8) - (10). 6
8 As a general conclusion we conjecture that a 3-stage estimator is preferable than a 2-stage one. Furthermore, if the DGP is characterized by strict monotonicity with high derivsative, Smooth-FDH should be considered, but if that is not the case Max-Smooth is preferable Simulation II Frontier I: Here we consider also FDH and DEA estimators. It is worth noting that the DGP considered in this case is very favorable to DEA. Even though Smooth-FDH estimator outperforms its competitors in terms of MSE and almost all situations presented in Figures (11) - (14). Frontier II: In this simulation Max-Smooth shows the best performance among all estimators considered in terms of MSE and in almost all situations considered presented on Figures (15) - (18). It is worth noting that in this set of simulations (Simulation II) we have to estimate two bandwidths in order to implement Max-Smooth estimator, whereas just one bandwidth to implement Smooth-FDH. Even though Max-Smooth exhibits a better performance. This shows that selecting two bandwidths is worthing the price in this case. As a result the conclusions from Simulation II are quite similar to those presented for Simulation I. Furthermore, selecting two bandwidths is not a big concern since the relative performance between Smooth-FDH and Max-Smooth seems not to be affected if we compare Simulations I and II. 5 Conclusion In this paper we propose two estimators for deterministic production frontier models named as Max-Smooth and Smooth-FDH. The referred estimators have desirable properties: (i) they are smooth; (ii) they are not inherent biased as FDH and DEA estimators, and (iii) they are simple to implement. We also present a cross-validation procedure for bandwidth selection. We compare the relative performance of the proposed estimators with DEA and FDH in a simulation study. The results are very favorable to the proposed estimators even in a simulation setup that is very suitable for DEA estimator. Although the results are very satisfactory some extensions are desirable. In a future work we intend to establish a bandwidth selection criteria for the 2-stage estimator (MaxS-2S) and to analyze the performance of that estimator in the presence of outliers. We conjecture that MaxS-2S is very appropriate in that case. 7
9 Figura 3: Frontier I - MSE of Frontier Estimators - Simulation I Figura 4: Frontier I - Dispersion of Frontier Estimators around x 1 = Simulation I Figura 5: Frontier I - Dispersion of Frontier Estimators around x 2 = Simulation I Figura 6: Frontier I - Dispersion of Frontier Estimators around x 3 = Simulation I
10 Figura 7: Frontier II - MSE of Frontier Estimators - Simulation I Figura 8: Frontier II - Dispersion of Frontier Estimators around x 1 = Simulation I Figura 9: Frontier II - Dispersion of Frontier Estimators around x 2 = Simulation I Figura 10: Frontier II - Dispersion of Frontier Estimators around x 3 = Simulation I
11 Figura 11: Frontier I - MSE of Frontier Estimators - Simulation II DEA DEA DEA Figura 12: Frontier I - Dispersion of Frontier Estimators around x 1 = Simulation II DEA DEA DEA Figura 13: Frontier I - Dispersion of Frontier Estimators around x 2 = Simulation II DEA DEA DEA Figura 14: Frontier I - Dispersion of Frontier Estimators around x 3 = Simulation II DEA DEA DEA
12 Figura 15: Frontier II - MSE of Frontier Estimators - Simulation II Figura 16: Frontier II - Dispersion of Frontier Estimators around x 1 = Simulation II Figura 17: Frontier II - Dispersion of Frontier Estimators around x 2 = Simulation II Figura 18: Frontier II - Dispersion of Frontier Estimators around x 3 = Simulation II
13 6 References Aigner, D., C.A.K. Lovell and P. Schmidt, 1977, Formulation and estimation of stochastic frontiers production function models. Journal of Econometrics 6, Cazals, C., J.-P. Florens and L. Simar, 2002, Nonparametric frontier estimation: a robust approach. Journal of Econometrics 106, Charnes, A.,W. Cooper and E. Rhodes, 1978, Measuring the efficiency of decision making units. European Journal of Operational Research 2, Deprins, D., L. Simar and H. Tulkens, 1984, Measuring labor inefficiency in post offices, in: M. Marchand, P. Pestiau and H. Tulkens, (Eds.), The performance of public enterprises: concepts and measurements. North Holland, Amsterdam. Fan, J., 1992, Design-adaptive Nonparametric Regression. Journal of the American Statistical Association, Vol. 87, No. 420, Fan, J. and I. Gijbels, 1995, Data driven bandwidth selection in local polynomial fitting: variable bandwidth and spatial adaptation. Journal of the Royal Statistical Society, B 57, Fan, J. and Gijbels, I., 1996, Local Polynomial Modelling and Its Applications. London: Chapman and Hall. Fan, J., and Q. Yao, 1998, Efficient estimation of conditional variance functions in stochastic regression. Biometrika 85, Farrell, M., 1957, The measurement of productive efficiency. Journal of the Royal Statistical Society A 120, Gijbels, I., E. Mammen, B. Park and L. Simar, 1999, On estimation of monotone and concave frontier functions. Journal of the American Statistical Association 94, Korostelev, A. P., L. Simar and A. B. Tsybakov, 1995, Efficient estimation of monotone boundaries. Annals of Statistics 23, Martins-Filho, C. and Yao, F., 2007, Nonparametric frontier estimation via local linear regression. Econometrics. Journal of Park, B., L. Simar and Ch. Weiner, 2000, The FDH estimator for productivity efficient scores: asymptotic properties. Econometric Theory 16, Seifford, L., 1996, Data envelopment analysis: the evolution of the state of the art ( ). Journal of Productivity Analysis 7,
Nonparametric Frontier estimation: A Multivariate Conditional Quantile Approach
Nonparametric Frontier estimation: A Multivariate Conditional Quantile Approach Abdelaati Daouia and Léopold Simar GREMAQ, Université de Toulouse I allée de Brienne 3 TOULOUSE, France (e-mail: daouia@cict.fr)
More informationNonparametric Efficiency Analysis: A Multivariate Conditional Quantile Approach
Nonparametric Efficiency Analysis: A Multivariate Conditional Quantile Approach Abdelaati Daouia GREMAQ, Université de Toulouse I and LSP, Université de Toulouse III Abdelaati.Daouia@math.ups-tlse.fr Léopold
More informationNonparametric regression using kernel and spline methods
Nonparametric regression using kernel and spline methods Jean D. Opsomer F. Jay Breidt March 3, 016 1 The statistical model When applying nonparametric regression methods, the researcher is interested
More informationNonparametric Estimation of Distribution Function using Bezier Curve
Communications for Statistical Applications and Methods 2014, Vol. 21, No. 1, 105 114 DOI: http://dx.doi.org/10.5351/csam.2014.21.1.105 ISSN 2287-7843 Nonparametric Estimation of Distribution Function
More informationConditional Nonparametric Frontier Models for Convex and Non Convex Technologies: a Unifying Approach
Conditional Nonparametric Frontier Models for Convex and Non Convex Technologies: a Unifying Approach Cinzia Daraio IIT-CNR and Scuola Superiore S. Anna, Italy cinzia@sssup.it, cinzia.daraio@iit.cnr.it
More informationFree Disposal Hull(FDH) Analysis for Efficiency Measurement: An update to dea
The Stata Journal (yyyy) vv, Number ii, pp. 1 8 Free Disposal Hull(FDH) Analysis for Efficiency Measurement: An update to dea Byoungin Lim Kyoungrok Lee Nextree Soft Co., Ltd. Korea National Defense University
More informationIntroduction to Nonparametric/Semiparametric Econometric Analysis: Implementation
to Nonparametric/Semiparametric Econometric Analysis: Implementation Yoichi Arai National Graduate Institute for Policy Studies 2014 JEA Spring Meeting (14 June) 1 / 30 Motivation MSE (MISE): Measures
More informationConditional Volatility Estimation by. Conditional Quantile Autoregression
International Journal of Mathematical Analysis Vol. 8, 2014, no. 41, 2033-2046 HIKARI Ltd, www.m-hikari.com http://dx.doi.org/10.12988/ijma.2014.47210 Conditional Volatility Estimation by Conditional Quantile
More informationRobust Shape Retrieval Using Maximum Likelihood Theory
Robust Shape Retrieval Using Maximum Likelihood Theory Naif Alajlan 1, Paul Fieguth 2, and Mohamed Kamel 1 1 PAMI Lab, E & CE Dept., UW, Waterloo, ON, N2L 3G1, Canada. naif, mkamel@pami.uwaterloo.ca 2
More informationLocally Weighted Least Squares Regression for Image Denoising, Reconstruction and Up-sampling
Locally Weighted Least Squares Regression for Image Denoising, Reconstruction and Up-sampling Moritz Baecher May 15, 29 1 Introduction Edge-preserving smoothing and super-resolution are classic and important
More information4.12 Generalization. In back-propagation learning, as many training examples as possible are typically used.
1 4.12 Generalization In back-propagation learning, as many training examples as possible are typically used. It is hoped that the network so designed generalizes well. A network generalizes well when
More informationThe Cross-Entropy Method
The Cross-Entropy Method Guy Weichenberg 7 September 2003 Introduction This report is a summary of the theory underlying the Cross-Entropy (CE) method, as discussed in the tutorial by de Boer, Kroese,
More informationNonparametric and Semiparametric Econometrics Lecture Notes for Econ 221. Yixiao Sun Department of Economics, University of California, San Diego
Nonparametric and Semiparametric Econometrics Lecture Notes for Econ 221 Yixiao Sun Department of Economics, University of California, San Diego Winter 2007 Contents Preface ix 1 Kernel Smoothing: Density
More informationNonparametric Survey Regression Estimation in Two-Stage Spatial Sampling
Nonparametric Survey Regression Estimation in Two-Stage Spatial Sampling Siobhan Everson-Stewart, F. Jay Breidt, Jean D. Opsomer January 20, 2004 Key Words: auxiliary information, environmental surveys,
More informationChapter 6: Examples 6.A Introduction
Chapter 6: Examples 6.A Introduction In Chapter 4, several approaches to the dual model regression problem were described and Chapter 5 provided expressions enabling one to compute the MSE of the mean
More informationChapter 3. Bootstrap. 3.1 Introduction. 3.2 The general idea
Chapter 3 Bootstrap 3.1 Introduction The estimation of parameters in probability distributions is a basic problem in statistics that one tends to encounter already during the very first course on the subject.
More informationExploring Econometric Model Selection Using Sensitivity Analysis
Exploring Econometric Model Selection Using Sensitivity Analysis William Becker Paolo Paruolo Andrea Saltelli Nice, 2 nd July 2013 Outline What is the problem we are addressing? Past approaches Hoover
More informationAn Empirical Comparison of Spectral Learning Methods for Classification
An Empirical Comparison of Spectral Learning Methods for Classification Adam Drake and Dan Ventura Computer Science Department Brigham Young University, Provo, UT 84602 USA Email: adam drake1@yahoo.com,
More informationA Fast Clustering Algorithm with Application to Cosmology. Woncheol Jang
A Fast Clustering Algorithm with Application to Cosmology Woncheol Jang May 5, 2004 Abstract We present a fast clustering algorithm for density contour clusters (Hartigan, 1975) that is a modified version
More informationA noninformative Bayesian approach to small area estimation
A noninformative Bayesian approach to small area estimation Glen Meeden School of Statistics University of Minnesota Minneapolis, MN 55455 glen@stat.umn.edu September 2001 Revised May 2002 Research supported
More informationAn algorithm for censored quantile regressions. Abstract
An algorithm for censored quantile regressions Thanasis Stengos University of Guelph Dianqin Wang University of Guelph Abstract In this paper, we present an algorithm for Censored Quantile Regression (CQR)
More informationAssessing the Quality of the Natural Cubic Spline Approximation
Assessing the Quality of the Natural Cubic Spline Approximation AHMET SEZER ANADOLU UNIVERSITY Department of Statisticss Yunus Emre Kampusu Eskisehir TURKEY ahsst12@yahoo.com Abstract: In large samples,
More informationON SWELL COLORED COMPLETE GRAPHS
Acta Math. Univ. Comenianae Vol. LXIII, (1994), pp. 303 308 303 ON SWELL COLORED COMPLETE GRAPHS C. WARD and S. SZABÓ Abstract. An edge-colored graph is said to be swell-colored if each triangle contains
More information10-701/15-781, Fall 2006, Final
-7/-78, Fall 6, Final Dec, :pm-8:pm There are 9 questions in this exam ( pages including this cover sheet). If you need more room to work out your answer to a question, use the back of the page and clearly
More informationThe Ohio State University Columbus, Ohio, USA Universidad Autónoma de Nuevo León San Nicolás de los Garza, Nuevo León, México, 66450
Optimization and Analysis of Variability in High Precision Injection Molding Carlos E. Castro 1, Blaine Lilly 1, José M. Castro 1, and Mauricio Cabrera Ríos 2 1 Department of Industrial, Welding & Systems
More informationA popular method for moving beyond linearity. 2. Basis expansion and regularization 1. Examples of transformations. Piecewise-polynomials and splines
A popular method for moving beyond linearity 2. Basis expansion and regularization 1 Idea: Augment the vector inputs x with additional variables which are transformation of x use linear models in this
More informationLEAST DISTANCE BASED INEFFICIENCY MEASURES ON THE PARETO-EFFICIENT FRONTIER IN DEA
Journal of the Operations Research Society of Japan Vol. 55, No. 1, March 2012, pp. 73 91 c The Operations Research Society of Japan LEAST DISTANCE BASED INEFFICIENCY MEASURES ON THE PARETO-EFFICIENT FRONTIER
More informationOn the Parameter Estimation of the Generalized Exponential Distribution Under Progressive Type-I Interval Censoring Scheme
arxiv:1811.06857v1 [math.st] 16 Nov 2018 On the Parameter Estimation of the Generalized Exponential Distribution Under Progressive Type-I Interval Censoring Scheme Mahdi Teimouri Email: teimouri@aut.ac.ir
More informationParametric versus non-parametric simulation
Parametric versus non-parametric simulation Bérénice Dupeux 1, Jeroen Buysse 2 1 Faculty of Bioscience Engineering, Ghent university, Coupure Links 653, Gent 9000, Belgium 2 Faculty of Bioscience Engineering,
More informationThe Simple Genetic Algorithm Performance: A Comparative Study on the Operators Combination
INFOCOMP 20 : The First International Conference on Advanced Communications and Computation The Simple Genetic Algorithm Performance: A Comparative Study on the Operators Combination Delmar Broglio Carvalho,
More informationLOGISTIC REGRESSION FOR MULTIPLE CLASSES
Peter Orbanz Applied Data Mining Not examinable. 111 LOGISTIC REGRESSION FOR MULTIPLE CLASSES Bernoulli and multinomial distributions The mulitnomial distribution of N draws from K categories with parameter
More informationNonparametric Regression
Nonparametric Regression John Fox Department of Sociology McMaster University 1280 Main Street West Hamilton, Ontario Canada L8S 4M4 jfox@mcmaster.ca February 2004 Abstract Nonparametric regression analysis
More informationChapter 7: Dual Modeling in the Presence of Constant Variance
Chapter 7: Dual Modeling in the Presence of Constant Variance 7.A Introduction An underlying premise of regression analysis is that a given response variable changes systematically and smoothly due to
More informationActive Network Tomography: Design and Estimation Issues George Michailidis Department of Statistics The University of Michigan
Active Network Tomography: Design and Estimation Issues George Michailidis Department of Statistics The University of Michigan November 18, 2003 Collaborators on Network Tomography Problems: Vijay Nair,
More informationTHE LINEAR PROGRAMMING APPROACH ON A-P SUPER-EFFICIENCY DATA ENVELOPMENT ANALYSIS MODEL OF INFEASIBILITY OF SOLVING MODEL
American Journal of Applied Sciences 11 (4): 601-605, 2014 ISSN: 1546-9239 2014 Science Publication doi:10.3844/aassp.2014.601.605 Published Online 11 (4) 2014 (http://www.thescipub.com/aas.toc) THE LINEAR
More informationRobust Signal-Structure Reconstruction
Robust Signal-Structure Reconstruction V. Chetty 1, D. Hayden 2, J. Gonçalves 2, and S. Warnick 1 1 Information and Decision Algorithms Laboratories, Brigham Young University 2 Control Group, Department
More informationGeneralized Additive Model
Generalized Additive Model by Huimin Liu Department of Mathematics and Statistics University of Minnesota Duluth, Duluth, MN 55812 December 2008 Table of Contents Abstract... 2 Chapter 1 Introduction 1.1
More informationSandeep Kharidhi and WenSui Liu ChoicePoint Precision Marketing
Generalized Additive Model and Applications in Direct Marketing Sandeep Kharidhi and WenSui Liu ChoicePoint Precision Marketing Abstract Logistic regression 1 has been widely used in direct marketing applications
More informationMotivation. Technical Background
Handling Outliers through Agglomerative Clustering with Full Model Maximum Likelihood Estimation, with Application to Flow Cytometry Mark Gordon, Justin Li, Kevin Matzen, Bryce Wiedenbeck Motivation Clustering
More informationThe strong chromatic number of a graph
The strong chromatic number of a graph Noga Alon Abstract It is shown that there is an absolute constant c with the following property: For any two graphs G 1 = (V, E 1 ) and G 2 = (V, E 2 ) on the same
More informationGuidelines for the application of Data Envelopment Analysis to assess evolving software
Short Paper Guidelines for the application of Data Envelopment Analysis to assess evolving software Alexander Chatzigeorgiou Department of Applied Informatics, University of Macedonia 546 Thessaloniki,
More information3 No-Wait Job Shops with Variable Processing Times
3 No-Wait Job Shops with Variable Processing Times In this chapter we assume that, on top of the classical no-wait job shop setting, we are given a set of processing times for each operation. We may select
More informationA Trimmed Translation-Invariant Denoising Estimator
A Trimmed Translation-Invariant Denoising Estimator Eric Chicken and Jordan Cuevas Florida State University, Tallahassee, FL 32306 Abstract A popular wavelet method for estimating jumps in functions is
More informationSketchable Histograms of Oriented Gradients for Object Detection
Sketchable Histograms of Oriented Gradients for Object Detection No Author Given No Institute Given Abstract. In this paper we investigate a new representation approach for visual object recognition. The
More informationMultivariate Analysis
Multivariate Analysis Project 1 Jeremy Morris February 20, 2006 1 Generating bivariate normal data Definition 2.2 from our text states that we can transform a sample from a standard normal random variable
More informationRecent advances in Metamodel of Optimal Prognosis. Lectures. Thomas Most & Johannes Will
Lectures Recent advances in Metamodel of Optimal Prognosis Thomas Most & Johannes Will presented at the Weimar Optimization and Stochastic Days 2010 Source: www.dynardo.de/en/library Recent advances in
More informationFEAR 1.14 User s Guide
FEAR 1.14 User s Guide Paul W. Wilson Department of Economics 222 Sirrine Hall Clemson University Clemson, South Carolina 29634 USA pww@clemson.edu 10 October 2010 This manual is for FEAR version 1.14,
More informationFuzzy satisfactory evaluation method for covering the ability comparison in the context of DEA efficiency
Control and Cybernetics vol. 35 (2006) No. 2 Fuzzy satisfactory evaluation method for covering the ability comparison in the context of DEA efficiency by Yoshiki Uemura Faculty of Education Mie University,
More informationGeneral properties of staircase and convex dual feasible functions
General properties of staircase and convex dual feasible functions JÜRGEN RIETZ, CLÁUDIO ALVES, J. M. VALÉRIO de CARVALHO Centro de Investigação Algoritmi da Universidade do Minho, Escola de Engenharia
More informationComparing different interpolation methods on two-dimensional test functions
Comparing different interpolation methods on two-dimensional test functions Thomas Mühlenstädt, Sonja Kuhnt May 28, 2009 Keywords: Interpolation, computer experiment, Kriging, Kernel interpolation, Thin
More informationInclusion of Aleatory and Epistemic Uncertainty in Design Optimization
10 th World Congress on Structural and Multidisciplinary Optimization May 19-24, 2013, Orlando, Florida, USA Inclusion of Aleatory and Epistemic Uncertainty in Design Optimization Sirisha Rangavajhala
More informationDiscount curve estimation by monotonizing McCulloch Splines
Discount curve estimation by monotonizing McCulloch Splines H.Dette, D.Ziggel Ruhr-Universität Bochum Fakultät für Mathematik 44780 Bochum, Germany e-mail: holger.dette@ruhr-uni-bochum.de Februar 2006
More informationCLASSIFICATION WITH RADIAL BASIS AND PROBABILISTIC NEURAL NETWORKS
CLASSIFICATION WITH RADIAL BASIS AND PROBABILISTIC NEURAL NETWORKS CHAPTER 4 CLASSIFICATION WITH RADIAL BASIS AND PROBABILISTIC NEURAL NETWORKS 4.1 Introduction Optical character recognition is one of
More informationNon-Parametric and Semi-Parametric Methods for Longitudinal Data
PART III Non-Parametric and Semi-Parametric Methods for Longitudinal Data CHAPTER 8 Non-parametric and semi-parametric regression methods: Introduction and overview Xihong Lin and Raymond J. Carroll Contents
More informationToday. Lecture 4: Last time. The EM algorithm. We examine clustering in a little more detail; we went over it a somewhat quickly last time
Today Lecture 4: We examine clustering in a little more detail; we went over it a somewhat quickly last time The CAD data will return and give us an opportunity to work with curves (!) We then examine
More informationAdvanced Operations Research Techniques IE316. Quiz 1 Review. Dr. Ted Ralphs
Advanced Operations Research Techniques IE316 Quiz 1 Review Dr. Ted Ralphs IE316 Quiz 1 Review 1 Reading for The Quiz Material covered in detail in lecture. 1.1, 1.4, 2.1-2.6, 3.1-3.3, 3.5 Background material
More informationImproving the Post-Smoothing of Test Norms with Kernel Smoothing
Improving the Post-Smoothing of Test Norms with Kernel Smoothing Anli Lin Qing Yi Michael J. Young Pearson Paper presented at the Annual Meeting of National Council on Measurement in Education, May 1-3,
More informationHOUGH TRANSFORM CS 6350 C V
HOUGH TRANSFORM CS 6350 C V HOUGH TRANSFORM The problem: Given a set of points in 2-D, find if a sub-set of these points, fall on a LINE. Hough Transform One powerful global method for detecting edges
More informationRegularization and model selection
CS229 Lecture notes Andrew Ng Part VI Regularization and model selection Suppose we are trying select among several different models for a learning problem. For instance, we might be using a polynomial
More informationLecture 2: Introduction to Numerical Simulation
Lecture 2: Introduction to Numerical Simulation Ahmed Kebaier kebaier@math.univ-paris13.fr HEC, Paris Outline of The Talk 1 Simulation of Random variables Outline 1 Simulation of Random variables Random
More informationORT EP R RCH A ESE R P A IDI! " #$$% &' (# $!"
R E S E A R C H R E P O R T IDIAP A Parallel Mixture of SVMs for Very Large Scale Problems Ronan Collobert a b Yoshua Bengio b IDIAP RR 01-12 April 26, 2002 Samy Bengio a published in Neural Computation,
More informationGradient LASSO algoithm
Gradient LASSO algoithm Yongdai Kim Seoul National University, Korea jointly with Yuwon Kim University of Minnesota, USA and Jinseog Kim Statistical Research Center for Complex Systems, Korea Contents
More informationAnalysis of high dimensional data via Topology. Louis Xiang. Oak Ridge National Laboratory. Oak Ridge, Tennessee
Analysis of high dimensional data via Topology Louis Xiang Oak Ridge National Laboratory Oak Ridge, Tennessee Contents Abstract iii 1 Overview 1 2 Data Set 1 3 Simplicial Complex 5 4 Computation of homology
More informationLecture 2 - Introduction to Polytopes
Lecture 2 - Introduction to Polytopes Optimization and Approximation - ENS M1 Nicolas Bousquet 1 Reminder of Linear Algebra definitions Let x 1,..., x m be points in R n and λ 1,..., λ m be real numbers.
More informationCOPYRIGHTED MATERIAL. Introduction. Chapter 1
Chapter 1 Introduction Performance Analysis, Queuing Theory, Large Deviations. Performance analysis of communication networks is the branch of applied probability that deals with the evaluation of the
More informationSOM+EOF for Finding Missing Values
SOM+EOF for Finding Missing Values Antti Sorjamaa 1, Paul Merlin 2, Bertrand Maillet 2 and Amaury Lendasse 1 1- Helsinki University of Technology - CIS P.O. Box 5400, 02015 HUT - Finland 2- Variances and
More informationLecture 2 September 3
EE 381V: Large Scale Optimization Fall 2012 Lecture 2 September 3 Lecturer: Caramanis & Sanghavi Scribe: Hongbo Si, Qiaoyang Ye 2.1 Overview of the last Lecture The focus of the last lecture was to give
More informationSplines and penalized regression
Splines and penalized regression November 23 Introduction We are discussing ways to estimate the regression function f, where E(y x) = f(x) One approach is of course to assume that f has a certain shape,
More information15.10 Curve Interpolation using Uniform Cubic B-Spline Curves. CS Dept, UK
1 An analysis of the problem: To get the curve constructed, how many knots are needed? Consider the following case: So, to interpolate (n +1) data points, one needs (n +7) knots,, for a uniform cubic B-spline
More informationSpatial Patterns Point Pattern Analysis Geographic Patterns in Areal Data
Spatial Patterns We will examine methods that are used to analyze patterns in two sorts of spatial data: Point Pattern Analysis - These methods concern themselves with the location information associated
More informationA Cumulative Averaging Method for Piecewise Polynomial Approximation to Discrete Data
Applied Mathematical Sciences, Vol. 1, 16, no. 7, 331-343 HIKARI Ltd, www.m-hiari.com http://dx.doi.org/1.1988/ams.16.5177 A Cumulative Averaging Method for Piecewise Polynomial Approximation to Discrete
More informationHomework. Gaussian, Bishop 2.3 Non-parametric, Bishop 2.5 Linear regression Pod-cast lecture on-line. Next lectures:
Homework Gaussian, Bishop 2.3 Non-parametric, Bishop 2.5 Linear regression 3.0-3.2 Pod-cast lecture on-line Next lectures: I posted a rough plan. It is flexible though so please come with suggestions Bayes
More informationA Geometric Analysis of Subspace Clustering with Outliers
A Geometric Analysis of Subspace Clustering with Outliers Mahdi Soltanolkotabi and Emmanuel Candés Stanford University Fundamental Tool in Data Mining : PCA Fundamental Tool in Data Mining : PCA Subspace
More informationINFO0948 Fitting and Shape Matching
INFO0948 Fitting and Shape Matching Renaud Detry University of Liège, Belgium Updated March 31, 2015 1 / 33 These slides are based on the following book: D. Forsyth and J. Ponce. Computer vision: a modern
More informationADAPTIVE METROPOLIS-HASTINGS SAMPLING, OR MONTE CARLO KERNEL ESTIMATION
ADAPTIVE METROPOLIS-HASTINGS SAMPLING, OR MONTE CARLO KERNEL ESTIMATION CHRISTOPHER A. SIMS Abstract. A new algorithm for sampling from an arbitrary pdf. 1. Introduction Consider the standard problem of
More informationSpatial Outlier Detection
Spatial Outlier Detection Chang-Tien Lu Department of Computer Science Northern Virginia Center Virginia Tech Joint work with Dechang Chen, Yufeng Kou, Jiang Zhao 1 Spatial Outlier A spatial data point
More informationAn ICA based Approach for Complex Color Scene Text Binarization
An ICA based Approach for Complex Color Scene Text Binarization Siddharth Kherada IIIT-Hyderabad, India siddharth.kherada@research.iiit.ac.in Anoop M. Namboodiri IIIT-Hyderabad, India anoop@iiit.ac.in
More informationA response to the critiques of DEA by Dmitruk and Koshevoy, and Bol
J Prod Anal (2008) 29:15 21 DOI 10.1007/s11-007-0061-7 A response to the critiques of DEA by Dmitruk and Koshevoy, and Bol W. W. Cooper Æ Z. Huang Æ S. Li Æ J. Zhu Published online: 1 August 2007 Ó Springer
More informationOn the Computational Complexity of Nash Equilibria for (0, 1) Bimatrix Games
On the Computational Complexity of Nash Equilibria for (0, 1) Bimatrix Games Bruno Codenotti Daniel Štefankovič Abstract The computational complexity of finding a Nash equilibrium in a nonzero sum bimatrix
More informationNonparametric Approaches to Regression
Nonparametric Approaches to Regression In traditional nonparametric regression, we assume very little about the functional form of the mean response function. In particular, we assume the model where m(xi)
More informationAlmost Curvature Continuous Fitting of B-Spline Surfaces
Journal for Geometry and Graphics Volume 2 (1998), No. 1, 33 43 Almost Curvature Continuous Fitting of B-Spline Surfaces Márta Szilvási-Nagy Department of Geometry, Mathematical Institute, Technical University
More informationText Modeling with the Trace Norm
Text Modeling with the Trace Norm Jason D. M. Rennie jrennie@gmail.com April 14, 2006 1 Introduction We have two goals: (1) to find a low-dimensional representation of text that allows generalization to
More informationDATA DEPTH AND ITS APPLICATIONS IN CLASSIFICATION
DATA DEPTH AND ITS APPLICATIONS IN CLASSIFICATION Ondrej Vencalek Department of Mathematical Analysis and Applications of Mathematics Palacky University Olomouc, CZECH REPUBLIC e-mail: ondrej.vencalek@upol.cz
More informationQUANTIZER DESIGN FOR EXPLOITING COMMON INFORMATION IN LAYERED CODING. Mehdi Salehifar, Tejaswi Nanjundaswamy, and Kenneth Rose
QUANTIZER DESIGN FOR EXPLOITING COMMON INFORMATION IN LAYERED CODING Mehdi Salehifar, Tejaswi Nanjundaswamy, and Kenneth Rose Department of Electrical and Computer Engineering University of California,
More informationOn Kernel Density Estimation with Univariate Application. SILOKO, Israel Uzuazor
On Kernel Density Estimation with Univariate Application BY SILOKO, Israel Uzuazor Department of Mathematics/ICT, Edo University Iyamho, Edo State, Nigeria. A Seminar Presented at Faculty of Science, Edo
More informationThe cointardl addon for gretl
The cointardl addon for gretl Artur Tarassow Version 0.51 Changelog Version 0.51 (May, 2017) correction: following the literature, the wild bootstrap does not rely on resampled residuals but the initially
More informationAn Improved Upper Bound for the Sum-free Subset Constant
1 2 3 47 6 23 11 Journal of Integer Sequences, Vol. 13 (2010), Article 10.8.3 An Improved Upper Bound for the Sum-free Subset Constant Mark Lewko Department of Mathematics University of Texas at Austin
More informationDigital Image Processing Laboratory: MAP Image Restoration
Purdue University: Digital Image Processing Laboratories 1 Digital Image Processing Laboratory: MAP Image Restoration October, 015 1 Introduction This laboratory explores the use of maximum a posteriori
More informationMOTION OF A LINE SEGMENT WHOSE ENDPOINT PATHS HAVE EQUAL ARC LENGTH. Anton GFRERRER 1 1 University of Technology, Graz, Austria
MOTION OF A LINE SEGMENT WHOSE ENDPOINT PATHS HAVE EQUAL ARC LENGTH Anton GFRERRER 1 1 University of Technology, Graz, Austria Abstract. The following geometric problem originating from an engineering
More informationNatural Quartic Spline
Natural Quartic Spline Rafael E Banchs INTRODUCTION This report describes the natural quartic spline algorithm developed for the enhanced solution of the Time Harmonic Field Electric Logging problem As
More informationLocal Polynomial Order in Regression Discontinuity Designs 1
Local Polynomial Order in Regression Discontinuity Designs 1 David Card UC Berkeley, NBER and IZA David S. Lee Princeton University and NBER Zhuan Pei Brandeis University Andrea Weber University of Mannheim
More informationData Mining Chapter 3: Visualizing and Exploring Data Fall 2011 Ming Li Department of Computer Science and Technology Nanjing University
Data Mining Chapter 3: Visualizing and Exploring Data Fall 2011 Ming Li Department of Computer Science and Technology Nanjing University Exploratory data analysis tasks Examine the data, in search of structures
More information16.410/413 Principles of Autonomy and Decision Making
16.410/413 Principles of Autonomy and Decision Making Lecture 17: The Simplex Method Emilio Frazzoli Aeronautics and Astronautics Massachusetts Institute of Technology November 10, 2010 Frazzoli (MIT)
More informationAN APPROXIMATE INVENTORY MODEL BASED ON DIMENSIONAL ANALYSIS. Victoria University, Wellington, New Zealand
AN APPROXIMATE INVENTORY MODEL BASED ON DIMENSIONAL ANALYSIS by G. A. VIGNAUX and Sudha JAIN Victoria University, Wellington, New Zealand Published in Asia-Pacific Journal of Operational Research, Vol
More informationChapter 15 Introduction to Linear Programming
Chapter 15 Introduction to Linear Programming An Introduction to Optimization Spring, 2015 Wei-Ta Chu 1 Brief History of Linear Programming The goal of linear programming is to determine the values of
More informationSection 4 Matching Estimator
Section 4 Matching Estimator Matching Estimators Key Idea: The matching method compares the outcomes of program participants with those of matched nonparticipants, where matches are chosen on the basis
More informationNONPARAMETRIC REGRESSION WIT MEASUREMENT ERROR: SOME RECENT PR David Ruppert Cornell University
NONPARAMETRIC REGRESSION WIT MEASUREMENT ERROR: SOME RECENT PR David Ruppert Cornell University www.orie.cornell.edu/ davidr (These transparencies, preprints, and references a link to Recent Talks and
More informationKernel Density Estimation (KDE)
Kernel Density Estimation (KDE) Previously, we ve seen how to use the histogram method to infer the probability density function (PDF) of a random variable (population) using a finite data sample. In this
More informationPerformance Limitations of Some Industrial PID Controllers
Performance Limitations of Some ndustrial P Controllers Flávio Faccin and Jorge O. Trierweiler * Chemical Engineering epartment Federal University of Rio Grande do Sul, Porto Alegre - RS, Brazil Abstract
More informationDATA MINING AND MACHINE LEARNING. Lecture 6: Data preprocessing and model selection Lecturer: Simone Scardapane
DATA MINING AND MACHINE LEARNING Lecture 6: Data preprocessing and model selection Lecturer: Simone Scardapane Academic Year 2016/2017 Table of contents Data preprocessing Feature normalization Missing
More information