Institute for Mathematical Sciences Event Archive
Online registration form
Enquiries |
|
Empirical Likelihood Based Methods in Statistics
(6 June - 1 July 2016)
Organizing Committee · Visitors and Participants · Overview · Activities · Venue
- Sanjay Chaudhuri (National University of Singapore)
- Song Xi Chen (Peking University and Iowa State University)
- Malay Ghosh (University of Florida)
- Ian McKeague (Columbia University)
- Art Owen (Stanford University)
- Cheng Yong Tang (Temple University)
Empirical likelihood based methods are becoming more and more popular in current statistics and econometrics. It is a semi-parametric method which allows the user to specify a parameter based model through estimating equations. However, there is no need to specify any distribution for data generation. This distribution is estimated from the data by a constrained empirical estimate. Information about the parameter is included through the constraints imposed by the estimating equations.
Owen [1988] was the first to use empirical likelihood as an alternative to likelihood ratio tests and derived its asymptotic distribution. His seminal work together with the advent powerful computational facilities have made empirical likelihood based method immensely popular among the practitioners. Estimation of parameters by maximising empirical likelihoods were first studied by Qin and Lawless [1994]. It is seen that these estimates possess many nice properties we associate with fully parametric maximum likelihood estimates.
There are several advantages of using empirical likelihood based methods. It is usually easier to specify and handle estimating equations than a full fledged parametric model. Empirical likelihood is easy to compute. Under the true model the empirical likelihood based estimates are almost as efficient to their parametric counterparts. However, if the parametric distribution for data generation is misspecified empirical likelihood based estimates are often more efficient. These properties have lead to an increased popularity of empirical likelihood based methods in severely constrained problems. Examples include statistical analysis with missing data, clinical trials, surrogate data, two phase designs, data augmentation from several sources, demographic applications, covariance estimation, graphical Markov models, analysis of survival data with various form of censoring, analysis of complex survey data etc.
The programme would start with a week long tutorial on the basics of empirical likelihood and its applications. There will be two workshops on the second and the third weeks. The first workshop would explore various aspects of the new devel-opments in the core empirical likelihood methodology. It would concentrate more on the theoretical side. As for example, we would discuss asymptotic behaviour of empirical likelihood under various levels of dependence, its behaviour in high dimensions, in various constrained problems, theoretical aspects of Bayesian empirical likelihood, etc. The second workshop will mostly be on the applications of empirical likelihood in real life problems. The topics would include the applications in biostatistics, survival analysis, in the analysis of complex survey data, longitudinal data, finance and econometrics among others. The fourth week is kept free for research interactions. Most of it would be spent in informal discussions.
- Tutorial on Empirical Likelihood: 6 − 10 June 2016
- Workshop on Recent Developments in Empirical Likelihood Methodology: 13 − 17 June 2016
- Workshop on New Applications of Empirical Likelihood: 20 − 24 June 2016
- Collaborative Research: 27 June - 1 July 2016
Students and researchers who are interested in attending these activities are requested to complete the online registration form.
The following do not need to register:
- Those invited to participate.
Organizing Committee · Visitors and Participants · Overview · Activities · Venue
