On the use of marginal and conditional likelihood method for estimating the inflated parameter in zero inflated poisson regression model

Guddattu, Vasudeva and Rao, Aruna K (2012) On the use of marginal and conditional likelihood method for estimating the inflated parameter in zero inflated poisson regression model. ProbStat Forum, 5. pp. 54-61. ISSN 0974-3235

[img] PDF
one_of_the_use_of_marginal----.pdf - Published Version
Restricted to Registered users only

Download (115kB) | Request a copy

Abstract

In this paper we investigate utility of conditional and marginal likelihood approach for estimating inflation parameter form zero inflated Poisson regression. We consider a hospital based prospective study in which factors associated with number of adverse drug reactions is of interest. Further number of Adverse Drug Reactions (ADR) depends on length of stay in hospital. In first method we consider length of stay as an o�set variable and estimate estimates of inflation parameter and factors associated with Number of ADR. In second method we consider length of stay as a random e�ect which follows exponential distribution and estimated estimate of inflation parameter, factors associated with mean response by considering marginal distribution of mean response. The marginal distribution turns out to be zero inflated geometric distribution. Further simulation study was performed to investigate the validity of the exponential distribution assumptions. Results from simulation study suggest that estimation of coverage probability and average length of confidence interval for inflation parameter by condition likelihood method performs better than marginal likelihood methods.

Item Type: Article
Subjects: Departments at MU > Statistics
Depositing User: KMC Manipal
Date Deposited: 24 Aug 2012 04:18
Last Modified: 24 Aug 2012 04:18
URI: http://eprints.manipal.edu/id/eprint/77068

Actions (login required)

View Item View Item