Graduate Seminar

Location:  CL 312

Speaker: Himansi Kumari

MSc Student supervised by Dr. James H. McVittie

Title: Exploring the EM Algorithm: Examples and Length-Biased Data

Abstract:

A useful procedure in statistical inference, particularly for datasets with missing or incomplete information, is the Expectation-Maximation (EM) algorithm. We will illustrate how to apply the EM algorithm to classical and modern examples with specific focus on the algorithm’s implementation from initialization to convergence. Our modern example will be based on non-parametric estimation of the failure time survival function for length-biased right-censored data, a common type of data found in epidemiology. We will show how the EM algorithm accounts for the features of length-bias and censoring when estimating the unknown probability mass parameters.