Event Date & TimeClick the calendar icon to add events to your Personal Calendar
Important!
Please note that adding this event to your calendar does not mean you are registered to this event. Please make sure to complete your registration for this event below.
|
Location
UM Bannatyne: Chown Building: Room 207
Address:
Description:
Latent class models are a powerful, flexible approach to finding and describing otherwise hidden and unknown subpopulations. They can provide key insights for etiology discovery and patient-oriented treatment when “one size” models are inadequate. Advances in statistical methodology and computational power have made it a viable tool of growing popularity. Applications include topics such as market segmentation, gambling and substance abuse risks, clinical depression, and emergency medicine.
Objectives:
- Provide an introduction to latent class analysis from an applied perspective;
- Demonstrate the PROC LCA routine in SAS and related R packages.
By the end of the workshop, participants will be able to compare LCA with related models; interpret LCA output and make conclusions from it; understand the limitations of traditional LCA; and employ extensions of LCA to include covariates, parameter constraints, and distal outcomes.
Documents:
Cost:
$50 – Academic – Researchers and Students
Requirements:
Experience with latent modeling is not required, however knowledge of generalized linear regression modeling and categorical data analysis is recommended. Some familiarity with SAS or R may be an asset, but is not required.
Session 1, Monday, October 29
Foundational concepts of traditional LCA will be gently discussed, including the mathematical model and its estimation routines. The interpretation of parameter estimates and the comparison of competing models will be of focus. PROC LCA will be demonstrated alongside the applied examples. An optional hands-on session will be run for those who bring laptops with either SAS PROC LCA or R installed.
Session 2, Tuesday, October 30
The basic LCA model will be extended to include covariates and distal outcomes, as well as the investigation of parameter invariance across multiple groups of interest. Time permitting, other advanced topics and computer programs may be discussed. An optional hands-on session will be run for those who bring laptops.
Instructor:
Brenden Dufault, MSc, is a full-time biostatistical consultant with the George and Fay Yee Centre for Healthcare Innovation with degrees in statistics and epidemiology. He has experience with the design and analysis of studies in clinical medicine, microbiology, immunology, public health, nursing, and psychology. Although a generalist, some of his specializations include mixed-effects models, latent mixture models, and high-dimensional data.
Registration Cancellation Policy:
A registration refund will be made upon written request on or before October 21, 2018. A $35 administrative fee will be retained. No refunds will be made for cancellations after this date.