On testing the missing at random assumption (2006)
| Venue: | In Proceedings of the 17th European Conference on Machine Learning (ECML-2006 |
| Citations: | 12 - 0 self |
BibTeX
@INPROCEEDINGS{Jaeger06ontesting,
author = {Manfred Jaeger and Aalborg Universitet and Fredrik Bajers Vej E and Dk- Aalborg Ø},
title = {On testing the missing at random assumption},
booktitle = {In Proceedings of the 17th European Conference on Machine Learning (ECML-2006},
year = {2006},
pages = {671--678}
}
Bookmark
 |
 |
 |
 |
 |
 |
OpenURL
Abstract
Abstract. Most approaches to learning from incomplete data are based on the assumption that unobserved values are missing at random (mar). While the mar assumption, as such, is not testable, it can become testable in the context of other distributional assumptions, e.g. the naive Bayes assumption. In this paper we investigate a method for testing the mar assumption in the presence of other distributional constraints. We present methods to (approximately) compute a test statistic consisting of the ratio of two profile likelihood functions. This requires the optimization of the likelihood under no assumptions on the missingness mechanism, for which we use our recently proposed AI & M algorithm. We present experimental results on synthetic data that show that our approximate test statistic is a good indicator for whether data is mar relative to the given distributional assumptions. 1
Citations
| 170 | D: Inference and missing data - Rubin - 1976 |
| 40 | Partial Identification of Probability Distributions - Manski - 2003 |
| 31 | Coarsening at random: characterizations, conjectures and counter-examples - Gill, Laan, et al. - 1997 |
| 19 | Ignorability for categorical data - Jaeger |
| 9 | Likelihood and Bayesian inference from selectively reported data - Dawid, Dickey - 1977 |
| 5 | The AI & M Procedure for Learning from Incomplete data - Jaeger - 2006 |
| 1 | Ignorability and coarse data. The Annals of Statistics - Heitjan, Rubin |
