Register now to access 7 million high quality study materials (What's Course Hero?) Course Hero is the premier provider of high quality online educational resources. With millions of study documents, online tutors, digital flashcards and free courseware, Course Hero is helping students learn more efficiently and effectively. Whether you're interested in exploring new subjects or mastering key topics for your next exam, Course Hero has the tools you need to achieve your goals.

16 Pages

Course: ECON 350, Fall 2009
School: Concordia Chicago
Rating:

Word Count: 1163

Document Preview

Note A on Adapting Propensity Score Matching and Selection Models to Choice Based Samples James J. Heckman and Petra E. Todd November 2004 This draft, February 16, 2006 1 Introduction The probability of selection into a treatment, or the propensity score, plays a central role in classical selection models and in matching methods (see, e.g., Heckman (1980), Rosenbaum and Rubin (1983)). Heckman and Robb (1986)...

Register Now

Unformatted Document Excerpt

Coursehero >> Illinois >> Concordia Chicago >> ECON 350

Course Hero has millions of student submitted documents similar to the one
below including study guides, practice problems, reference materials, practice exams, textbook help and tutor support.

Course Hero has millions of student submitted documents similar to the one below including study guides, practice problems, reference materials, practice exams, textbook help and tutor support.
Note A on Adapting Propensity Score Matching and Selection Models to Choice Based Samples James J. Heckman and Petra E. Todd November 2004 This draft, February 16, 2006 1 Introduction The probability of selection into a treatment, or the propensity score, plays a central role in classical selection models and in matching methods (see, e.g., Heckman (1980), Rosenbaum and Rubin (1983)). Heckman and Robb (1986) discuss how the propensity score is used di erently in matching and selection models. They also show that, given the propensity score, both matching and selection models are robust to choice based sampling, which occurs when treatment group members are over or underrepresented relative to their frequency in the population. This type of sampling design is frequently chosen in evaluation studies to reduce the costs of data collection and to get a larger number of observations on treated individuals. Given a consistent estimate of the propensity score, matching and clas2 sical selection methods are robust to choice-based sampling, because both are "control function" methods and are defined conditional on treatment and comparison group status. This note extends Heckman and Robb's (1986) analysis to consider the case where population weights are unknown so that propensity score cannot be consistently estimated. In evaluation settings, the population weights are often unknown or cannot easily estimated.1 For example, for the NSW training program studied in LaLonde (1986), the population consists of all persons eligible for the program, which was targeted at drug-addicts, ex-convicts, and welfare recipients. Few datasets The methods of Manski and Lerman (1977), Manski (1986) for adjusting for choice-based sampling in estimating the discrete choice probabilities cannot be applied when the weights are unknown and cannot be estimated from the data. 1 3 would have the information necessary to determine whether a person is eligible for the program, so it would be di cult to estimate the population weights needed to consistently estimate propensity scores. In this note, we establish that matching and selection procedures can still be applied when the propensity score is estimated on unweighted choice based samples. The idea is very simple. To implement both matching and classical selection models, only a monotonic transformation of the propensity score is required. In choice based samples, the odds ratio of the misspecified propensity score is monotonically related to the odds ratio of the true propensity scores. Thus selection and matching procedures can identify population treatment e ects using misspecified estimates of propensity scores fit on choice based samples. 4 2 Discussion of the Proposition Let = 1 if a person is a treatment group member; =0 if a member of the comparison group. = is a realization of . In the population generated from random sampling, the joint density is ( ) = [Pr( for = 1 | )] [Pr( = = 0 | )]1 {0 1} ( ) where is the density of the data. By Bayes's theorem, we have, letting Pr( = 1) = , ( | and ( | = 0)(1 ) = ( ) Pr( 5 =0| ) (1b) = 1) = ( ) ( =1| ) (1a) Take the ratio of (1a) to (1b), ( | = 1) Pr( = ( | = 0) 1 Pr( =1| ) =0| ) (2) Assume 0 Pr( = 1 | ) 1. From knowledge of the densities of the data in the two samples, ( | = 1) and ( | = 0) one can form a scalar multiple of the ratio of the propensity score without knowing . The odds ratio is a monotonic function of the propensity score that does not require knowledge of the true sample weights. In a choicebased sample, both the numerator and denominator of the first term in (2) can be consistently estimated. This monotonic function can replace ( ) in implementing both matching and nonparametric selection models. However, estimating ( | = ) is demanding of the data 6 when is of high dimension. Instead of estimating these densities, we can substitute for the left hand side of (2) the odds ratio of the estimated conditional probabilities obtained using the choice-based sample with the wrong weights. (i.e., ignoring the fact that the data are a choice based sample). The odds ratio of the estimated probabilities is a scalar multiple of the true odds ratio. It can therefore be used instead Pr( of = 1 | ) to match or construct nonparametric control functions in selection bias models. In the choice-based sample, let ~ ( = 1 | ) be the conditional probability that = 1 and be the unconditional 6= , the true popprobability of sampling = 1, where ulation proportion. The joint density of the data from the 7 sampled population is [ ( | = 1) ][ ( | = 0)(1 )]1 = 0) 1 ) Using (1a) and (1b) to solve for ( | one may write the data density as Pr( =1| ) ( ) Pr( = 1) and ( | =0| ) ( ) (1 (1 ) 8 ~ ( and ~ ( =1| )= Pr( ( | = 1) =1| ) ( ) + ( | = 0)(1 ) (3a) ) (3b) Under random sampling, the right-hand sides of (3a) and (3b) are the limits to which the choice-based probabilities converge. Taking the ratio of (3a) to (3b), assuming the latter is not zero, 1 Pr( = 0 | ) ( ) 1 =0| )= ( | = 1) + ( | = 0)(1 9 one obtains ~ ( =1| ) Pr( = 1 | ) 1 (4) = ~ ( =0| ) Pr( = 0 | ) 1 Thus, one can estimate the ratio of the propensity score up to scale (the scale is the product of the two terms on the righthand side of (4)). Instead of matching or estimating semiparametric selection models using Pr( = 1 | ) (as in, for example, Ahn and Powell (1993), Heckman, Ichimura, Smith and Todd (1998), and Powell (1999)), we can instead use the odds ratio of the estimate ~ ( = 1 | ), which is monotonically related to the true Pr( = 1 | ). In the case of a logit ( ), ( ) = exp( ) (1 + exp( )), the log of this ratio becomes ~ ( =1| ) ln = ~ ~ ( =0| ) 10 where the slope coe cients are the true values and the intercept ~ 0 = 0 + ( (1 )) + ((1 ) )), where 0 is the true value. In implementing nearest neighbor matching estimators, matching on the log odds ratio gives identical estimates to matching on the (unknown) Pr( = 1 | ), because the odds ratio preserves the ranking of the neighbors. In application of either matching or classical selection bias correction methods, one must account for the usual problems of using estimated log odds ratios instead of true values. 11 For discussion related to using estimated propensity scores, see Hahn (1998) Heckman, Ichimura, Smith and Todd (1998) Ichimura, Smith and Todd (1998) Imbens, Hirano and Ridder (2003) 12 References [1] Ahn, H. and J. Powell (1993), "Semiparametric Estimation of Censored Selection Models With Nonparametric Selection Mechanism," Journal of Econometrics, 58, 3-29. [2] Hahn, J. (1998), "On the Role of the Propensity Score in E cient Estimation of Average Treatment E ects," Econometrica, 66(2), 315-331. [3] Heckman, J., (1980), "Add...

Find millions of documents on Course Hero - Study Guides, Lecture Notes, Reference Materials, Practice Exams and more. Course Hero has millions of course specific materials providing students with the best way to expand their education.

Below is a small sample set of documents:

Concordia Chicago - CS - 347
Date: Tue, 16 Apr 2002 16:12:16 -0500 (CDT)From: Catalin Lucian Dumitrescu &lt;catalind@cs.uchicago.edu&gt;On Power-Law Relationships of the Internet Topology-1. State the main contribution (or focus?) The paper presents several observations abou
University of Texas - CAM - 397
CAM 397-Introduction to Mathematical Modeling Fall 2008Mon AUGUST X SEPTEMBER X Labor Day 8 Arbogast Wed 27 Carey Fri 29 Carey3Arbogast5Arbogast10 Arbogast 17 Carey 24 Elber 1 Dawson12 Arbogast 19 Carey 26 (Student) 3 Dawson15 Carey 22
UNE - BIO - 208
The Autonomic Nervous SystemLecture 17, Holes: Chapter 11The Autonomic Nervous SystemA.K.A Involuntary Nervous System General visceral motor System 2 divisions: Sympathetic Parasympathetic Differences to Somatic System Effectors Efferent pathway
CSU Northridge - HFBUS - 019
COURSE SYLLABUSBUSINESS LAW 308: BUSINESS ORGANIZATIONS, AGENCY, AND EMPLOYMENT LAW Professor Zucker Spring 2009Office: Telephone: JH 4235 (818) 677-3964 7:30 a.m. to 8:00 a.m. and 9:15 a.m. to 10:00 a.m. 3:00 p.m. to 4:00 p.m.Office Hours: Tuesd
CSU Northridge - HFBUS - 019
White Collar Crime Course Syllabus BLAW 495 (Special Topics)Professor Zucker Spring 2009Office: Telephone: E-mail: Class Hours: Ticket No.: Office Hours:BB 4235 (818) 677-3964 (Campus) bruce.zucker@csun.edu Tuesdays 4:00 p.m. to 6:45 p.m. 17716
Concordia Chicago - GALTON - 385
MATHEMATICS/STATISTICS 385 BROWNIAN MOTION &amp; STOCHASTIC CALCULUS PROBLEM SET 3Problem 1. Gamma and Dirichlet Processes. The standard Gamma process is the subordinator Ys such that for each s &gt; 0 the distribution of the random variable Ys is the Gam
BYU - IT - 327
IT 327: Lab 9 Digital Channel Testing: QAM Constellations, Eye PatternsObjectives Gain an understanding of QAM and PSK constellation diagrams using M-Level encoding and decoding. Create an eye pattern using a regular oscilloscope. Equipment Used T
CSU Northridge - HFBUS - 019
IUPUI - CPT - 220
Mathematical ExpectationQuantitative Analysis II Professor Bob Orr Copyright 2000 by Robert H. Orr and the Trustees of Indiana UniversityProbability DistributionMathematical function which describes the probabilities of possible events in a sam
IUPUI - CPT - 220
Writing to LearnCPT 220 Professor Robert H. OrrCopyright 2001 by Robert H. Orr and the Trustees of Indiana UniversityWhy Writing to Learn?sWriting is primarily an exercise in Logic Writing is a learning activity as well as a communication act
Concordia Chicago - TTIC - 20030612
Concordia Chicago - TTIC - 20030814
N.C. State - ST - 421
ST421 Exam 2 NAMEHughes-Oliver October 18, 1996Show ALL your work, along with JUSTIFICATION for the steps you take.1. 5 points Suppose Y is a continuous random variable with cumulative distribution function c.d.f. FY y and X is a discrete random
N.C. State - ST - 421
ST421 Exam 3 NAMEHughes-Oliver November 10, 1995Show ALL your work, along with JUSTIFICATION for the steps you take.1. 20 points An absent-minded professor does not remember which of her 12 keys will open her o ce door. If she tries at random an
IUPUI - CSCI - 230
Department of Computer and Information Science, School of Science, IUPUICSCI 230Introduction to Computers - LanguagesDale Roberts, Lecturer Computer Science, IUPUI E-mail: droberts@cs.iupui.eduComputer OrganizationA Typical Von-Neumann Archit
Concordia Chicago - BLF - 0218
Twin Gaussian Processes for Structured PredictionLiefeng Bo and Cristian Sminchisescu TTI-C, {blf0218,crismin}@tti-c.orgWe describe twin Gaussian processes (TGP), a structured prediction method that uses Gaussian processes (GP) priors for both cova
UCSD - AB - 171
National Oceanography Centre, SouthamptonCruise Report No.14 RRS Charles Darwin Cruise CD171 01 MAY 15 JUN 2005A Trans-Atlantic hydrographic section at 36NPrincipal ScientistE L McDonagh2007National Oceanography Centre, Southampton Univer
IUPUI - CSCI - 240
Department of Computer and Information Science, School of Science, IUPUIIntroduction to Visual ProgrammingDale Roberts, Lecturer Computer Science, IUPUI E-mail: droberts@cs.iupui.eduDale RobertsEvent-driven ProgrammingEvent-driven programmin
IUPUI - CSCI - 230
Department of Computer and Information Science, School of Science, IUPUICSCI 230Basic I/O scanf()Dale Roberts, LecturerDepartment of Computer and Information ScienceIUPUIDale RobertsFormatting Input with ScanfscanfInput formatting Capa
IUPUI - CSCI - 240
CSCI 240Concepts in Object-Oriented DesignDale Roberts, Lecturer Computer Science, IUPUI E-mail: droberts@cs.iupui.eduExample: Classroom Attending the lecture we have several individuals Wade - loves Chinese food George - an outdoorsman We
IUPUI - CSCI - 230
Department of Computer and Information Science, School of Science, IUPUIA First C ProgramDale Roberts, Lecturer Computer Science, IUPUI E-mail: droberts@cs.iupui.eduDale RobertsTry Your First C Program#include &lt;stdio.h&gt; main() Indicates a {
IUPUI - CSCI - 230
Department of Computer and Information Science, School of Science, IUPUICSCI 230Pointers IntroductionDale Roberts, Lecturer Computer Science, IUPUI E-mail: droberts@cs.iupui.eduDale RobertsWhat is PointerCS Dept. Location A: Highway Interse
IUPUI - CSCI - 230
Department of Computer and Information Science, School of Science, IUPUICSCI 230Functions Scope, Parameter Passing, Storage SpecifiersDale Roberts, Lecturer IUPUI droberts@cs.iupui.eduDale Roberts1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
IUPUI - CSCI - 240
Department of Computer and Information Science, School of Science, IUPUIInput/Output Streams File IODale Roberts, Lecturer Computer Science, IUPUI E-mail: droberts@cs.iupui.eduDale Roberts1File StreamsProvides high level support for file o
IUPUI - CSCI - 230
Department of Computer and Information Science, School of Science, IUPUICSCI 230Information Representation:Floating Point RepresentationDale Roberts, Lecturer IUPUI droberts@cs.iupui.eduDale RobertsFractional NumbersExamples: 456.7810 = 4
Carleton - GEO - 110
Geology 110 Fall 2006 IntroductionLittle Chicago Gravel QuarryAt a gravel quarry near Little Chicago (west of I-35) we will examine sedimentary deposits of the Des Moines lobe that formed during the Wisconsinan glaciation (about 14,000 years ago)
Concordia Chicago - STAT - 234
butterfat425710661664732714934761744653725657421573535602537405874791721849567468975
UCSD - CSE - 140
CS 140L Lecture 7 Transformation between Mealy and Moore MachinesProfessor CK Cheng CSE Dept. UC San DiegoOutlines Framework Procedure Example: Transform Example: State Diagram Example: Output SequenceTransform from Mealy to Moore MachineM
Concordia Chicago - STAT - 234
companyvendor0.500.790.580.710.900.821.170.821.140.731.250.770.750.721.220.790.740.720.800.91
Concordia Chicago - STAT - 234
xy491331081081109382119931191149813410611413196875215310111611412912097116110
Concordia Chicago - STAT - 234
calls3 0 3 4 9 1 6 2 2 5 7 6 4 2 2 4 1 03 1 3 3 4 2 1 3 3 2 3 2 5 2 1 0 2 4
Concordia Chicago - STAT - 234
force28.824.430.125.626.423.922.122.527.628.120.827.724.425.124.626.328.222.226.324.4
Concordia Chicago - STAT - 234
concentrationvoltage01.72 01.68 0 1.74 50 2.04 50 2.1150 2.17 75 2.40 75 2.32 75 2.33 100 2.91100 3.00 100 2.89 150 4.47 150 4.51 150 4.43200 6.67 200 6.66 200 6.57
Concordia Chicago - STAT - 24600
11.2944924.6194310.337202.18226317.244892.78281013.935366.96334315.124412.6113445.9950357.06774414.106749.9204910.371581.41412111.484049.74377210.437285.515324
Concordia Chicago - STAT - 24600
1622672711851116127831
Concordia Chicago - STAT - 22000
Difference in earning between men and women?VariablesExp - Experience (in years)Edu - Education (in years)Earn - Earnings (in \$1000)Sex - categorical variable:0 - male1 - female
Concordia Chicago - STAT - 22000
Data from Table 2.8, IPSVariables: Year: time Overseas: annual total return on overseas stocks in % US: annual total return on U.S. stocks in %
Concordia Chicago - STAT - 22000
Data from Table 2.11 in IPSFour data sets for exploring correlation and regressionData set B - variables:X explanatory variableY response
Concordia Chicago - STAT - 22000
Data from Table 2.11 in IPSFour data sets for exploring correlation and regressionData set A - variables:X explanatory variableY response
Concordia Chicago - STAT - 22000
Properties of EstimatorsStatistical properties of a and ^ ^ b Mean and variance of ^ bE(^ = b b)2 ^ = var(b) SXXRecall thatnSXX =i=1(xi - x)2 Distribution of ^ b2 ^ N b, b SXXMean and variance of a ^E(^) = a avar(^) = a x2 1 +
Concordia Chicago - STAT - 22000
Simple Linear RegressionExample: Body density Aim: Measure body density (weight per unit volume of the body) (Body density indicates the fat content of the human body.)Problem: Body density is difficult to measure directly. Research suggests tha
Concordia Chicago - STAT - 22000
Random VariablesAim: Learn about population Available information: observed data x1, . . . , xn Problem: Data aected by chance variation New set of data would look dierent Suppose we observe/measure some characteristic (variable) of n individual
Concordia Chicago - STAT - 24600
Statistics 24600 - Spring 2004 A heuristic argument for the limit of MLE when the model is misspecifiedKullback-Leibler distance between two density functions f and g is defined to be KL(f, g) Ef (log(f /g) =Rlog(f (x)/g(x)f (x)dx.Given i.i.d
Concordia Chicago - STAT - 24600
AdministriviaCourse Staff Lecturer Michael Eichler Eckhart 105 Course Assistant Xiaofeng Shao Meeting Times Lectures Section 1 TuTh 1:30-2:50 Discussion Session Eck 133 Course web pageAdministriviahttp:/www.stat.uchicago.edu/~eichler/stat24600 H
Concordia Chicago - WEB - 119
Physical Sciences 119 Stellar Astronomy and Astrophysics List of Course Activities 2006 Sept. 25 Lecturer: Donald G. York, AAC002, Phone 702 8930 don@oddjob.uchicago.edu My office is in the basement of the Enrico Fermi Institute. Enter at 5640 South
Carleton - PSYC - 124
Psych 124 Winter 2005Psych 124 Winter 2005The Pearson Correlation Coefficient Sum of Products (SP) of the deviations from the meanSP = &quot; ( xi ! x )( yi ! y )i=1 n nExample: Trailer Parks &amp; Tornados X = number of trailer parks in a city Y = nu
Stanford - SKIL - 1026
UNITED STATES DISTRICT COURT DISTRICT OF NEW HAMPSHIREIN RE SMARTFORCE PLC SECURITIES LITIGATION THIS DOCUMENT RELATES TO: ALL ACTIONS NO. 02-CV-544-B CHIEF JUDGE PAUL J. BARBADOROThe enclosed packet contains the following documents, which outline
Carleton - ECON - 111
Scott Bierman PROBLEM SET 2Principles of Microeconomics1. Carefully identify the opportunity cost of each of the following activities. a. What is the opportunity cost to you of purchasing a \$50 U.S. Savings Bond? b. What is the opportunity cost t
UCSD - COMPANY - 157
Designer:_ Production:_ Design Field:_ Company 157 Designer Contract Purpose: To set down basic guidelines of what is expected from the designer and Company 157 Agreement, which if violated, will be on record for public access. 1. I will call the s
Carleton - PHYS - 260
Amrit Tuladhar Physics 260: Final paper Materials for holographic data storage This is the age of information in more ways than one. First of all, more information than ever is being churned out by individuals and corporations around the world: we co
UCSD - SE - 253
SE 253B 1. Free VibrationHomework 6Due: 3/13 Not CollectedConsider the specially orthotropic plate having C-SS-SS-SS boundary conditions. Plate is [04/902]s carbon/epoxy laminate analyzed in previous class examples. Density is 0.06 lbm/in3. (N
SUNY Buffalo - CSE - 116