Advances and Applications in Statistics
Volume 46, Issue 2, Pages 97 - 105
(August 2015) http://dx.doi.org/10.17654/ADASAug2015_097_105 |
|
BOOTSTRAPPING METHOD FOR CONSTRUCTING CONFIDENTIAL INTERVALS FOR ESTIMATED PARAMETERS OF THE HIDDEN MARKOV MODEL
Yosuke Inaba and Etsuo Miyaoka
|
Abstract: The Hidden Markov Model (HMM) is a statistical model for analysing time series data where the observational data depends on unobservable states HMM is used in such fields as speech recognition and bioinformatics, and has recently gained popularity in life science as well. In HMM, we need to estimate the parameters, and constructing their confidential interval involves some difficulties. To solve this problem, we propose a bootstrapping method to construct confidential intervals and apply it to earthquake data. When applying this method, we need to choose an appropriate number of states, hence we calculated each AIC score, and then made comparisons. |
Keywords and phrases: Hidden Markov Model, HMM, bootstrap, likelihood-function, Baum-Welch, forward-backward algorithm. |
|
Number of Downloads: 402 | Number of Views: 1320 |
|