Multiple regression: bayesian inference
View/ Open
Date
2011Author
Omukami, Howard
Type
ThesisLanguage
en_USMetadata
Show full item recordAbstract
Bayesian approach is based on subjective interpretation of probability. It views probability as a degree of belief concerning an uncertainty. It.also regards an unknown parameter as an uncertainty on which a degree of belief can be expressed and then revised based on sample information. A parameter is viewed as a random variable which prior to sample evidence is assigned a prior distribution. When sample evidence is obtained, the prior distribution is revised and posterior distribution obtained.
The Bayesian approach also uses data to update the uncertainty distribution for unknown parameters then draw conclusions using the updated distributions. Bayesian inferences help in decision making such that the gains made out of a study are maximized and the risks minimized. Data is used to generate posterior distribution depending on assumed prior distribution. This is done in terms of a likelihood function corresponding to the observed data.
There are many situations especially in business, industry and technology where a careful analysis of a decision making is beneficial. .
In this project, the data analyzed was collected from a pig breeding research centre. Various factors affecting the weights of pigs were determined by a regression equation after computing various Bayesian point estimates.
Keywords
Parameters, Bayesian approach, minimum and maximum risk, posterior and priori distribution, likelihood function, regression, Bayesian point estimate.
Publisher
University of Nairobi, Kenya