Statistics Seminar - Stephen Walter - Evaluating classification accuracy without a gold standard: an interesting “Rule of Three”

Description

Title: Evaluating classification accuracy without a gold standard: an interesting “Rule of Three”

Speaker: Stephen Walter (McMaster University)

Abstract: 

For many diagnostic tests and other classifiers,  a reference or gold standard classification may not exist. Here, latent class modelling can be used to evaluate test accuracy for a variety of observational designs.  In this talk, I will focus on the minimal requirements of the data for model identifiability, contrasting the number of model parameters and the number of degrees of freedom available.  We will identify the smallest number of observations needed for identifiability in many situations (spoiler alert:there is a clue in the title). A particular focus will be on recent work involving replications of the same test method. Some practical examples will illustrate the approach.

Date/Time: Tuesday October 26, 2021, 3:30 - 4:30 

Location: Virtual 

Join Zoom Meeting
https://mcmaster.zoom.us/j/97199003250?pwd=dTVzUW5YaWovRm5GMEpwanpxT2JuZz09

Meeting ID: 971 9900 3250
Passcode: 643951
Go Back
McMaster University - Faculty of Science | Math & Stats