Variable Selection via MCMC Matching Pursuit
碩士 === 國立高雄大學 === 統計學研究所 === 95 === Usually Bayesian variable selection methods require computing an inversion of a p×p matrix at each iteration of the algorithms, where p is the number of the variables. However, this computational cost is very expensive, especially when p is larger and larger. In o...
| Main Authors: | , |
|---|---|
| Other Authors: | |
| Format: | Others |
| Language: | en_US |
| Online Access: | http://ndltd.ncl.edu.tw/handle/58898329264774919332 |
| Summary: | 碩士 === 國立高雄大學 === 統計學研究所 === 95 === Usually Bayesian variable selection methods require computing an inversion of a p×p matrix at each iteration of the algorithms, where p is the number of the variables. However, this computational cost is very expensive, especially when p is larger and larger. In order to avoid computing inverse matrices, two MCMC matching pursuit algorithms are modified here to be two variable selection procedures for solving variable selection problems. Several simulations and real examples are demonstrated here to show the performances of these two procedures.
|
|---|