IMS
Session Slot: 8:30-10:20 Thursday
Estimated Audience Size:
AudioVisual Request: Two Overheads
Session Title: Quasi Likelihood and Semiparametric Inference
Theme Session: No
Applied Session: No
Session Organizer: Li, Bing The Pennsylvania State University
Address: Department of Statistics The Pennsylvania State University 326 Joab Thomas Building University Park, PA 16802, USA
Phone: (814) 865 1952
Fax: (814) 863 7114
Email: bing@stat.psu.edu
Session Timing: 110 minutes total (Sorry about format):
Opening Remarks by Chair - 5 or 0 minutes First Speaker - 30 minutes Second Speaker - 30 minutes Third Speaker - 30 minutes Discussant - Floor Discusion - 10 minutes (or 5 or 15)
Session Chair: Shen, Xiaotong Ohio State University
Address: Department of Statistics, Ohio State University
Phone:
Fax:
Email:
1. Projection, Hedging, and Pricing of Derivatives
McLeish, Donald, University of Waterloo
Address: Department of Statistics and Actuarial Science University of Waterloo Waterloo, Ontario, Canada N2L 3G1
Phone: (519) 888-4567, ext. 5534
Fax: (519) 664-1762
Email: dlmcleis@setosa.uwaterloo.ca
Kolkiewicz, Adam, University of Waterloo
Abstract: The use of projection techniques to hedge risk and price derivative securities is discussed. This permits either a semi-parametric or a non-parametric specification of models and approximate hedging, designed, for example, to limit the costs associated with continuously rebalancing a portfolio.
2. A Projected Quasi-Score Method to Reducing Sensitivity to Nuisance Parameters: Application to Measurement Error Models
Rathouz, Paul J., University of Chicago
Address: University of Chicago Dept. of Health Studies 5841 S. Maryland Ave., MC2007 Chicago, IL 60637
Phone: 773-834-1970 (o)
Fax: 773-701-1979
Email: rathouz@galton.uchicago.edu
Liang, Kung-Yee, Johns Hopkins University
Abstract: The projection method proposed by Waterman and Lindsay (1996) for desensitization to nuisance parameters was originally based on the likelihood function. An analagous method is developed for a quasi-likelihood regression set-up, simplifying model specification, assumptions, and the development of software for data analyses with many nuisance parameters in the mean function. The degree to which the full likelihood method extends to the quasi-likelihood framework - and in what ways it falls short - is examined using finite sample properties of estimating functions. One application extends the work of Stefanski and Carroll (1987) to a broader class of measurement error models.
3. On Quasilikelihood Equations with Nonparametric Weights
Li, Bing, The Pennsylvania State University
Address: Department of Statistics The Pennsylvania State University 326 Joab Thomas Building University Park, PA 16802, USA
Phone: (814) 865 1952
Fax: (814) 863 7114
Email: bing@stat.psu.edu
Abstract: To apply the quasi likelihood method one needs to know the mean and the variance function, so as to give more weight to the more accurate observations. If the variance function is unknown, then one should acquire the knowledge of the weights from the sample. One way to do so is by adaptive estimation, which in this context involves the consistent estimation of the entire variance function. However, for moderate sample sizes, such attempt may render the estimate too unstable, and we need a more flexible method which does not involve consistent estimation of the variance but still gives optimal weights to the extent to which a moderate sample size can tolerate. In this work we introduce a new type of quasi likelihood equation, whose weights belong to a smooth class that need not be large enough to be adaptive, but which are nevertheless optimal among the specified class. We will also present a rigorous development of the large sample properties of this equation, together with a sufficient and necessary condition for adaptivity.
List of speakers who are nonmembers: None