Parametric and semi-parametric approaches to missing data

Placeholder Show Content


In this dissertation, I consider models that are designed to deal with missing data, in two very different areas of application, forecast combination and longitudinal studies. In both these areas, naive imputation methods can be shown to be justified only under the most restrictive and unrealistic assumptions, while going beyond naive imputation is challenging as it requires including more variables in the model, and so introduces nuisance parameters. Two distinct approaches to overfitting in the nuisance parameters are applied here. The empirical Bayes approach treats model parameters as random variables, and shrinks them towards some prior mean. In the first chapter, I apply the empirical Bayes approach to forecast combination. Using this approach to combine fore- casts from the Survey of Professional Forecasters, I find that the empirical Bayes model outperforms other models for short forecast horizons, but there is no clear result for longer forecast horizons. This can be attributed to the model failing to include serial correlation that exists in the data. In the second chapter, I consider a semiparametric approach to longitudinal studies where patients drop out at various points in time, creating a monotone missing data pattern. The semiparametric approach uses a variable number of parameters, in this case controlled by a choice of series order, so that the number of parameters in the model can vary with the size of the data set. In addition to a semiparametric estimator, we provide data driven method for the selection of the series order. This improvement, combined with other theoretical and practical advantages of our model, make it a practical alternative for empirical researchers, who have so far preferred likelihood based models that make very restrictive parametric assumptions.


Type of resource text
Form electronic; electronic resource; remote
Extent 1 online resource.
Publication date 2013
Issuance monographic
Language English


Associated with Tong, Kester
Associated with Stanford University, Department of Economics.
Primary advisor Hong, Han
Thesis advisor Hong, Han
Thesis advisor Romano, Joseph D
Thesis advisor Wolak, Frank A
Advisor Romano, Joseph D
Advisor Wolak, Frank A


Genre Theses

Bibliographic information

Statement of responsibility Kester Tong.
Note Submitted to the Department of Economics.
Thesis Thesis (Ph.D.)--Stanford University, 2013.
Location electronic resource

Access conditions

© 2013 by Kester Christopher Tong
This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).

Also listed in

Loading usage metrics...