Location

Pattillo Student Center, 2nd Floor

Start Date

29-4-2015 10:00 AM

End Date

29-4-2015 4:00 PM

Description

In this project, we develop a particular statistical model for binary data that allows for the possibility of false-positive misclassification. To account for the misclassification, the model incorporates a two-stage sampling scheme.

• Next, we apply maximum likelihood methods to find estimators of the primary prevalence parameter p as well as the false-positive misclassification rate parameter ϕ. In addition, we derive confidence intervals for p based on inverting Wald, score and likelihood ratio statistics.

• Also, we graphically compare coverage and width properties of the Wald-based, score-based, and likelihood ratio-based confidence intervals for p through a Monte Carlo simulation. The simulation study is done under different parameter and sample size configurations. Also, we apply the newly-derived confidence intervals for p to a real data set.

Included in

Mathematics Commons

Share

COinS

Tell us how this article helped you.

 
Apr 29th, 10:00 AM Apr 29th, 4:00 PM

A Fixed-Inverse Binary Misclassification Model Under Possible False-Positive Misclassification

Pattillo Student Center, 2nd Floor

In this project, we develop a particular statistical model for binary data that allows for the possibility of false-positive misclassification. To account for the misclassification, the model incorporates a two-stage sampling scheme.

• Next, we apply maximum likelihood methods to find estimators of the primary prevalence parameter p as well as the false-positive misclassification rate parameter ϕ. In addition, we derive confidence intervals for p based on inverting Wald, score and likelihood ratio statistics.

• Also, we graphically compare coverage and width properties of the Wald-based, score-based, and likelihood ratio-based confidence intervals for p through a Monte Carlo simulation. The simulation study is done under different parameter and sample size configurations. Also, we apply the newly-derived confidence intervals for p to a real data set.