|
Activity Number:
|
165
- SLDS CSpeed 2
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 10, 2021 : 10:00 AM to 11:50 AM
|
|
Sponsor:
|
Section on Statistical Learning and Data Science
|
|
Abstract #318529
|
|
|
Title:
|
Factor Analysis of Data with Incomplete Records
|
|
Author(s):
|
Fan Dai* and Somak Dutta and Ranjan Maitra
|
|
Companies:
|
Michigan Technological University and Iowa State University and Iowa State University
|
|
Keywords:
|
Incomplete data;
Factor model;
Profile likelihood;
Lanczos algorithm;
L-BFGS-B;
GRBs
|
|
Abstract:
|
Data that are partially observed arise naturally in scientific fields such as astronomy and social science. Traditional methods for analyzing partially recorded data involves incomplete observation deletion or missing value imputation. We develop a CEM (Conditional Expectation Maximization) algorithm under the missing at random (MAR) assumption and further assume a factor covariance structure which can explain the variation in large data set using a few latent factors. We propose a novel matrix-free profile likelihood approach to estimate the covariance parameters with high computational efficiency. Our method is applied to analyze the variability in Gamma Ray Bursts (GRBs).
|
Authors who are presenting talks have a * after their name.
