2009Unpublished venueRequires access

Existence of maximum likelihood estimation for three-parameter log-normal distribution

Yincai Tang, Xiaoling Wei

Open publisher page 2 citations

Abstract

In the literature concerning maximum likelihood estimation (MLE), likelihood is always defined as the product of corresponding density function for the observations. When the distribution function is log-normal, the unboundedness problem will lead to nonexistence of the estimation. In this article, we introduce an observed error term to modify the traditional likelihood function and consider the condition for the existence of the MLEs of three parameters in the corrected likelihood function. It is shown through an example that the MLEs of three parameters exists if the condition is satisfied and the estimation is not sensitive to the choice of the observation error.

About this research paper

What this paper is about

In the literature concerning maximum likelihood estimation (MLE), likelihood is always defined as the product of corresponding density function for the observations. When the distribution function is log-normal, the unboundedness problem will lead to nonexistence of the estimation. In this article, we introduce an observed error term to modify the traditional likelihood function and consider the condition for the existence of the MLEs of three parameters in the corrected likelihood function. It is shown through an example that the MLEs of three parameters exists if the condition is satisfied and the estimation is not sensitive to the choice of the observation error.

Why it matters

OpenAlex reports 2 citations for this work. Citation counts describe recorded attention and do not establish research quality.

Key contribution

A contribution statement is not available in the OpenAlex record.

Method / approach

Method details are not available in the OpenAlex metadata.

Main findings

Findings are not separately available in the OpenAlex metadata.

Limitations

Limitations are not available in the OpenAlex metadata.

Applications

Application details are not available in the OpenAlex metadata.

Available abstract

In the literature concerning maximum likelihood estimation (MLE), likelihood is always defined as the product of corresponding density function for the observations. When the distribution function is log-normal, the unboundedness problem will lead to nonexistence of the estimation. In this article, we introduce an observed error term to modify the traditional likelihood function and consider the condition for the existence of the MLEs of three parameters in the corrected likelihood function. It is shown through an example that the MLEs of three parameters exists if the condition is satisfied and the estimation is not sensitive to the choice of the observation error.

Key concepts: Maximum likelihood, Likelihood function, Mathematics, Estimation theory, Statistics, Function (biology), Estimation, Maximum likelihood sequence estimation

Related papers

Back to paper searchBrowse research topicsOriginal source
Existence of maximum likelihood estimation for three-parameter log-normal distribution — Research Paper | ScholarLens