On testing the missing at random assumption

Publikation: Bidrag til bog/antologi/rapport/konference proceedingKonferenceartikel i proceedingForskningpeer review

14 Citationer (Scopus)


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 assumptionson 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.
TitelMachine Learning: ECML 2006 : 17th European Conference on Machine Learning. Berlin, Germany, September 2006. Proceedings
Antal sider8
StatusUdgivet - 2006
BegivenhedEuropean Conference on Machine Learning - Berlin, Tyskland
Varighed: 18 sep. 200622 sep. 2006


KonferenceEuropean Conference on Machine Learning

Bibliografisk note

Serie: Lecture Notes in Artificial Intelligence, Springer-Verlag, 4212, 0302-9743

Fingeraftryk Dyk ned i forskningsemnerne om 'On testing the missing at random assumption'. Sammen danner de et unikt fingeraftryk.