Missing observation analysis for matrixvariate time series data
Abstract
Bayesian inference is developed for matrixvariate dynamic linear models (MVDLMs), in order to allow missing observation analysis, of any subvector or submatrix of the observation time series matrix. We propose modifications of the inverted Wishart and matrix $t$ distributions, replacing the scalar degrees of freedom by a diagonal matrix of degrees of freedom. The MVDLM is then redefined and modifications of the updating algorithm for missing observations are suggested.
 Publication:

arXiv eprints
 Pub Date:
 May 2008
 arXiv:
 arXiv:0805.3831
 Bibcode:
 2008arXiv0805.3831T
 Keywords:

 Statistics  Methodology;
 Statistics  Applications
 EPrint:
 11 pages, 1 figure