The comparison between maximum likelihood estimation and Bayesian method: fitting to finite mixture model
<p>In the era of Big Data, statistical modelling plays important role in handling a</p><p>prodigious flow of datasets. The existing literatures regarding the performance of</p><p>maximum likelihood estimation and Bayesian method t...
Saved in:
| Main Author: | |
|---|---|
| Format: | thesis |
| Language: | eng |
| Published: |
2022
|
| Subjects: | |
| Online Access: | https://ir.upsi.edu.my/detailsg.php?det=9589 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | <p>In the era of Big Data, statistical modelling plays important role in handling a</p><p>prodigious flow of datasets. The existing literatures regarding the performance of</p><p>maximum likelihood estimation and Bayesian method that fit with finite mixture model</p><p>in time series modelling is still lacking. The main objective of this study was to compare</p><p>the maximum likelihood estimation and Bayesian method in fitting with finite mixture</p><p>model and determine the plausible method in analysing time series data. Also, this study</p><p>aimed to identify the number of components and the representation existed in time</p><p>series data. Additionally, this study also evaluated and modelled the exchange rate,</p><p>inflation rate, electrical and electronic export values in Malaysia, Thailand and the</p><p>Philippines using both methods that fit to finite mixture model. The finite mixture</p><p>model is an unsupervised learning model that can fit with all types of distributions and</p><p>hence modelling a variety of data. In this study, maximum likelihood estimation and</p><p>Bayesian method were adapted with finite mixture model to investigate the relationship</p><p>between sampled variables as both methods are well-known parameter estimation</p><p>method used in large sample study. As a result, the two components mixture model</p><p>obtained in sampled variables. Both approaches revealed that a negative relationship</p><p>presented between exchange rate with electrical and electronic export prices. Besides</p><p>that, a positive relationship exhibited between inflation rate with electrical and</p><p>electronic export prices. For exchange rate and inflation rate, negative relationship</p><p>occurred in the normal situation while no relationship existed in crisis period. In</p><p>conclusion, both methods provided almost similar results but maximum likelihood</p><p>estimation performed better than the Bayesian method. As an implication, the efficiency</p><p>of statistical method, importance of components representations and statistical</p><p>modelling highlighted in this study can be a guideline to statisticians who are interested</p><p>in the similar field.</p> |
|---|