Survey Working Group Spring 2012

This Week:

Abstract: Propensity score weighting adjustment is commonly used to handle nonresponse. When the response mechanism is non-ignorable in the sense that the response probability depends directly on the study variable, a follow-up sample is commonly used to obtain an unbi- ased estimator using the framework of two-phase sampling. In this two-phase sampling approach, the follow-up sample is assumed to re- spond completely. In practice, the follow-up sample is also subject to missingness. We consider the problem of propensity score weighting adjustment when there are several follow-ups and the final follow-up sample is also subject to missingness. We propose a novel approach, that makes use of the monotone structure of missing pattern and enables efficient estimation. The proposed method is more exible than the existing methods and can be directly applicable to complex survey sampling.

Abstract: Parameter estimation with missing data is frequently encountered problem in statistics. Imputation is often used to facilitate the parameter estimation by simply applying the complete-sample estimators to the imputed dataset.

We consider the problem of parameter estimation for longitudinal or Longitudinal data with missing values. We use the linear mixed effects models, in which we treat the unobserved random effects as “missing data” and consider a special case of the non-ignorable missing model that assumes the missingness depends on the missing values only through the random effect (Follmann&Wu1995). Our goal is to develop a fractional imputation method proposed by Kim (2011) under this response model. Fractional imputation simplifies the computation associated with the EM algorithm for maximum likelihood estimation with missing data. In the M-step, the restricted maximum likelihood method can be considered. Calibration fractional imputation is also considered as a way for improving the Monte Carlo approximation in the fractional imputation. Results from a limited simulation study are presented to compare the proposed method with the existing methods.
Slides

Senniang Chen: A Bayesian Analysis of a Jump-diffusion Model with double Exponential Jumps

Abstract: Li, Wells and Yu (2008) show that the jump-diffusion model (AJD) with compound Poisson jumps fail to capture the return dynamics, in particular, fail to capture the “infinite-activity” small jumps. A question raised recently in literature is whether a more complicated finite activity jump model, such as double exponential jumps proposed in Kou (2002), can adequately approximate the behavior of jumps in the return data. In this project, we develop a MCMC method for estimating the continuous-time model with stochastic volatility and double exponential jumps using discretely sampled data. Our simulation studies show that our MCMC method can provide accurate estimation of model parameters, stochastic volatility and jumps. However, our results from an empirical study show that the AJD model with double exponential jumps is inadequate of modeling S&P 500 return data from 1990-2010, indicating that infinite-activity jump models are still crucial for capturing the real jumps in return dynamics of S&P 500.

Seunghwan Park: Small area estimation incorporating information from several sources

Abstract: Combining information from different source is an important practical problem. The source of information can come from a probability sampling with direct measurement, from another probability sampling with indirect measurement, or from auxiliary area level information. We consider the area-level model approach to small area estimation with at least two survey information. The way we combine information is based on the generalized least squares estimation from the measurement error model, where the sampling error of the survey estimates of direct measurement can be treated as the measurement error. Mean square estimation is also discussed. The proposed method is applied to the Korean labor force survey problem.

Dr. Jae-kwang Kim: Combining data from two independent surveys: model-assisted approach

Abstract: Combining information from two or more independent surveys is a problem frequently encountered in survey sampling. We consider the case of two independent surveys, where a large sample from survey 1 collects only auxiliary information and a much smaller sample from survey 2 provides information on both the variables of interest and the auxiliary variables. We propose a model-assisted projection method of estimation based on a working model, but the reference distribution is design-based. We generate synthetic or proxy values of a variable of interest by first fitting the working model, relating the variable of interest to the auxiliary variables, to the data from survey 2 and then predict the variable of interest associated with the auxiliary variables observed in survey 1. The projection estimator of a total is simply obtained from the survey 1 weights and associated synthetic values. We identify the conditions for the projection estimator to be asymptotically unbiased. Domain estimation using the projection method is also considered. Replication variance estimators are obtained by augmenting the synthetic data file for survey 1 with additional synthetic columns associated with the columns of replicate weights. Results from a simulation study are presented. This is a joint work with J.N.K. Rao at Carleton University in Canada.