RIT on Weighted Estimating Equations
in Surveys and Biostatistics

Mon. 4-5,  Rm  1308                  Fall 2013 and continuing Spring 2014

Eric Slud        Statistics Program , Math Department        Rm 2314       x5-5469

                  Interested participants should send email to    evs@math.umd.edu

See the   Direction of emphasis for Spring 2014   from the org. meeting Mon. Feb. 3, 2014.

Reading list      updated for Spring 2014 with papers for Spring 2014 highlighted by red asterisks *

Schedule of Talks      updated with Slides where available

Research Focus: A great deal of current research in parametric, semiparametric and also
sample-survey statistical inference is organized around Estimating Equations. This includes

  • martingale estimating equations arising in Survival Analysis;
  • Inverse Probability Weighted estimating equations to enable estimation in biostatistics,
    survey sampling and other contexts where there is missing data, biased sampling, or nonresponse;
  • calibration methodology, empirical likelihood and other ways of efficiently
    incorporating auxiliary information into estimation;
  • quasilikelihood based estimating equations arising in Generalized Linear Models;
  • Generalized Estimating Equations (GEE's) arising in inference for longitudinal data;

    plus many other topics. We will study papers from a few of these areas, focusing in areas of
    interest to the RIT attendees.

    Graduate Student Prerequisites: To benefit from this research activity, a graduate student
    should have completed Stat 700-701 and Stat 600.

    Graduate Program: Graduate students will be involved in reading and presenting
    papers from the statistical literature concerning provable properties of estimators from
    Estimating Equations.

    Work Schedule: We will meet weekly. Students will choose from the following list of Topics
    and Papers (which will regularly be augmented on this web-page) and will present the material
    in subsequent weeks, after an introductory couple of weeks' talks. Presentations can be informal,
    but should be detailed enough and present enough background that we can understand
    the issues and ideas clearly. Some presentations will extend to a second week.

    Topics by Keyword:

  • misspecified regression models,
  • robustness under misspecifications, `double robustness'
  • random-effect GLM's,
  • errors-in-variables (`measurement error') models,
  • longitudinal models & GEE methods (Generalized Estimating Equations),
  • biased sampling & survey-weighted models
  • Also see material on previous web-pages concerning Semiparametric Satistics
    and statistics related to Biased Sampling.

    Reading List

    Chen, Jinbo and Norman Breslow (2004) Semiparametric efficient estimation for the auxiliary outcome
    problem with the conditional mean model,
    Canad. Jour. Statist. 32, 1-14.

    Chen, Jiahua and Qin, Jing (1993) Empirical Likelihood Estimation for Finite Populations and the Effective
    Usage of Auxiliary Information
    , Biometrika 80, 107-116.

    V. P. Godambe classic paper on optimal estimating equations,
             An Optimum Property of Regular Maximum Likelihood Estimation, pp. 1208-1211, Ann. Math. Stat. 31
    Stable URL: http://links.jstor.org/sici?sici=0003-4851%28196012%2931%3A4%3C1208%3AAOPORM%3E2.0.CO%3B2-K

    Godambe, V. and Thompson, M. (1986) Parameters of Superpopulation and Survey Population:
    Their Relationships and Estimation
    , International Statistical Review 54, 127-138.

    * Heyde, C. (1997), "Quasilikelihood and its Application", Springer book.

    Hirano, K., Imbens, G. and Ridder, G. (2003) Efficient estimation of average treatment effects using the
    estimated propensity score, Econometrica 71, 1161-1189.

    Huber, P. (1967) classic paper on M-estimation from the 5th Berkeley Symposium,
    The behavior of maximum likelihood estimates under nonstandard conditions,
    Proc. Fifth Berkeley Symp. on Math. Statist. and Prob., Vol. 1 (Univ. of Calif. Press, 1967), 221-233.

    Janicki, R. (2009) UMCP thesis on Estimating Equations including misspecified ones.

    Lumley, T., Shaw, P. and Dai, J. (2011), Connections between Survey Calibration Estimators and
    Semiparametric Models for Incomplete Data
    , International Statistical Review 79, 200-220.

    * Ma, Y. and Zhu, L. (2012), "A semiparametric approach to dimension reduction",
    Journal of American Statistical Association 107, 168-179.

    Pfeffermann, D. and Sverchkov, M.:   work on survey data with semiparametrically modelled
    informative nonresponse.

    * J. Robins papers (many with Rotnitzky and other authors) on inverse-probability weighted
    estimating equations, starting with

    * Robins, J., Rotnitzky, A. and Zhao, L. (1994), Estimation of regression coefficients when some regressors
    are not always observed
    , Jour. Amer. Statist. Assoc. 89, 846-866.

    * Tan, Z. several papers and discussions on missing data, causal inference, and double robustness, starting with:

    * Z. Tan (2007) Understanding OR, PS, and DR, Discussion of "Demystifying double robustness:
    A comparison of alternative strategies for estimating a population mean from incomplete
    data" by Kang and Schafer, Statistical Science 22, 560-568.

    * Tsiatis, A. (2006) book, "Semiparametric Theory and Missing Data", Springer.

    * Varin, C., Reid, N. and Firth, C. (2011), "An Overview of Composite Likelihood Methods" Statistica Sinica 21, 5-42.

    White, Halbert (1982) Maximum likelihood estimation of misspecified models.
             Econometrica 50, no. 1, 1-25.

    * Zeger, S., Liang, K and Albert, P. (1988) Models for longitudinal data: a generalized estimating equation approach,
    Biometrics 44, 1049-1060.

    Schedule of Talks ---

    Talks during Spring 2014

  • February 3, 2014:    Organizational meeting for Spring 2014: Introduction of the topic by Eric Slud.

  • February 10, 2014:    Xuan Yao presented theoretical background from the Tsiatis (2006) semiparametrics
    book, especially from Chapters 4 and 7, related to Regular Asymptotically Linear estimators, estimating equations,
    and influence functions
    in the setting of missing-data semiparametric problems of interest in this RIT.

  • February 17, 2014:    Xuan Yao will finish her presentation on material from Chapter 7 of the Tsiatis book, and
    Eric Slud will pick up where she leaves off to discuss the "influence functions" for optimal semiparametric
    estimators in outcome and response-propensity models. Slides can be found here. (This material involves
    Theorem statements from Chapters 8-10 of Tsiatis (2006), but after a quick statement of results, the rest of the
    presentation consists of working out examples.)

  • The RIT meeting for Feb. 24 was cancelled.

  • March 3: This meeting was cancelled because of snow.

  • March 10: Xia Li will talk about the Rotnizky, Robins and Zhao (1994, JASA) paper in the Reading List above.

  • March 17: Spring break.

  • March 24: Benjamin Kedem will speak on "Time Series Prediction by Out of Sample Fusion", about
    an original approach he has developed, along with students, to use imputed and augmented data in
    statistical inference.

  • March 31: Jong Jun Lee, speaking about Causal Inference using response propensities estimated
    nonparametrically, from the Hirano, Imbens, and Ridder (2003) paper in the Reading List above.

  • April 7: Eric Slud will speak about an Econometrica paper by J. Hahn (1998) on the role of
    propensity-score estimation in causal inference. This paper (by a student of Imbens) was cited in the
    2003 Econometrica paper of Hirano, Imbens and Ridder covered last week, and we will return to the
    discussion of that paper too.

  • April 14: Judith Law will speak on Causal Inference using Instrumental Variables (a 1996 JASA
    paper by Angrist, Imbens and Rubin).

  • April 21: This week's session was canceled.

  • April 28: There will be a meeting of the RIT on this date. The speaker is Xuan Yao
    who will talk about her recent reading in Arthur Owen's Empirical Likelihood book, to
    connect the general empirical likelihood theory to calibrated weighted estimating equation
    methods, including methods based on missing data.

  • May 5: Ruth Pfeiffer will speak about the Ma and Shu JASA 2012 paper on Semiparametric
    Dimension Reduction

  • May 12: Lemeng Pan will speak about Composite Likelihood methods of statistical inference,
    and their relation to the estimating equation topics we have studied.

  • Talks from Fall 2013

  • September 9, 2013:    Organizational meeting and Introductio of the topic by Eric Slud. Slides available
                                        here in form including Nov.18 presentation.

  • September 16, 2013:    Eric Slud, further informal introduction to connection between
                                        Survey Missing-Data problems and Estimating Equations.

  • September 23, 2013:    Meiyu Shen will speak about the Godambe (1960) Ann. Math.Stat. paper linked above.
                                        The general theme here is optimality within classes of estimating equations, which
                                        will be important throughout the RIT.

  • September 30, 2013:    Xuan Yao will speak about the H. White (1982) Econometrica paper linked above.
                                        Here the theme is misspecified models and auxiliary assumptions that may allow
                                        consistent estimation of parameters or testing within them.

  • October 7, 2013:    Continuation of Xuan Yao presentation of the Halbert White paper on misspecified models estimating equations.

  • October 14, 2013:    NOTE: this RIT session WILL meet; the departmental event that I thought would
                                        conflict will be held earlier in the afternoon.

                                           Jong Jun Lee will speak about Chapter 3 of the Tsiatis book, Semiparametric Theory and Missing Data.
                                        The theme is influence functions and asymptotically linear estimators in parametric estimation problems.

  • October 21, 2013:    Barry Graubard will present material relating Empirical Likelihood in survey problems to the weighted
                                        estimating equations topic, primarily the J. Chen and J. Qin Biometrika paper linked above (in the Reading
                                        list) via JSTOR. The theme is Empirical Likelihood as a way to make Survey/Semiparametrics connections.

  • October 28, 2013:    Barry Graubard will finish his presentation of the Chen and Qin empirical likelihood paper. Then
                                        Paul Smith will discuss the close connections between survey generalized regression estimators and
                                        calibration estimators, using as source the 1992 JASA paper of J.-C. Deville and C.-E. Sarndal.

  • November 4, 2013:    Meiyu Shen will speak about the Lumley et al. ISI Review paper linked above.
                                        The theme of this paper also is Survey/Semiparametrics connections.

  • November 11, 2013:    Meiyu Shen completed her presentation of the Lumley et al.paper.

  • November 18, 2013:    Eric Slud will speak about the semiparametric formulation of iid missing data problems with auxiliary
                                        information and how the form of such results might be made to relate to survey-nonresponse/calibration problems
                                        under superpopulation large-sample asymptotics. The most relevant readings are Chapters 4 and 6 of the Tsiatis
                                        book along with the Chen and Qin empirical-likelihood paper we head about on 10/21 and the Deville and Sarndal
                                        survey calibration paper we heard about on 10/28. This lecture is meant to consolidate some of the ideas we
                                        have been talking about, with the goal of introducing the research setting of the Z. Tan papers.

                                        Slides for the lecture are available here.

  • November 25, 2013:    Xia Li and Qi Liu will together present the Kang and Schafer (2007) Statistical Science paper mentioned in
                                        the reading list, along with the related ideas in the Z. Tan discussion of that paper. The topic is "Double Robustness".

  • © Last updated April 29, 2014.