Menu
Log in


Webinar: Random Effects Inference in Linear Mixed Models: The good, the bad, and the misspecified

  • 25 Sep 2020
  • 12:00 PM - 1:00 PM (UTC+10:00)
  • via Zoom
  • 11

Registration


Register

You are warmly invited to the following webinar hosted by SSA:

Random Effects Inference in Linear Mixed Models: The good, the bad, and the misspecified

held on Friday, 25 September 2020 at 12:00PM AEST via Zoom, exclusively for members of SSA.

This event is presented by Francis K.C. Hui and Alan H. Welsh (Research School of Finance, Actuarial Studies & Statistics, Australian National University).

About this webinar:

In many applications of mixed models, the scientific questions of interest can often relate to estimation and inference of the random effects . Examples range from assessing the importance of all or a subset of the variance components, constructing functions and associated uncertainty intervals of functions of the random effects and/or variance components, to directly modelling the random effects as a function of other covariates. On the other hand, unlike fixed effects inference there is a lot more contention and uncertainty (pun intended!) surrounding both how to do random effects inference in mixed models, and what the consequences are of associated misspecification on estimation and inference.

This talk is a culmination of two projects on the topic of random effects inference in linear mixed models. In the first half, we re-examine one of the earliest and simplest methods of random effects testing in linear mixed models, namely the F-test based on linear combinations of the responses, or FLC test. For current statistical practice, we argue that the FLC test is underused and should be given more consideration especially as an initial or “gateway” test for linear mixed models. We discuss three advantages of the FLC test often overlooked in modern applications of linear mixed models: computation speed, generality, and its exactness as a test. In the second half, we examine some impacts of random effects misspecification on random effects inference in linear mixed models. While some large sample results can be formulated, we show that in general that incorrectly assuming a normal distribution for the random effects can have severe consequences for random effects inference, with strongly biased estimation of variance components and associated likelihood-based confidence intervals for the variance components exhibited potentially severe under coverage.

If time permits, we conclude our talk by encouraging more general discussion and research on the notion of sometimes treating random effects as fixed effects for estimation and inference, and how the above findings may generalise far beyond linear mixed models.

About the presenters:

Francis K.C. Hui and Alan H. Welsh (Research School of Finance, Actuarial Studies & Statistics, Australian National University)

Francis Hui is a Senior Lecturer in Statistics at the Australian National University. He completed his PhD at the University of New South Wales in 2014 and moved to Canberra to undertake a postdoctoral fellowship at the ANU, and has been willingly stuck there since. His research spans a mixture of methodological, computational, and applied statistics, including longitudinal and correlated data analysis, dimension reduction and variable selection, and approximate statistical estimation and inference. Much of his research is motivated by joint modelling in ecology, and longitudinal analysis of mental health data, and is complemented by copious amounts of tea drinking, and unhealthy amounts of anime watching.

Alan Welsh, recipient of the Pitman Medal in 2012, is the E.J. Hannan Professor of Statistics at the Australian National University. He has been at ANU for a while, spending time in the Mathematical Sciences Institute and the Research School of Finance, Actuarial Studies & Statistics. He has also held positions at the University of Chicago and the University of Southampton. His research interests include statistical inference, statistical modelling, robustness, nonparametric and semiparametric methods, analysis of sample surveys, and ecological monitoring. Much of his recent work has gone into trying to understand linear mixed models better.

Registration

This event is for members of SSA only. It is free, but you will need to register. Registration is a 2-step process. Please use the registration link on the left to register with SSA. You will receive a confirmation email containing a link for the registration with Zoom.  Please complete the registration with Zoom at your earliest convenience as places on our Zoom platform are limited!

This event will be recorded and the recording added to SSA’s webinar page in due course after the event.

Would you please note that the times stated are AEST?

Cancellation Policy

Our Guarantee
Our goal is to provide truly exceptional offerings and service, and we won't be happy until you are. If any programs, products or services of SSA do not fulfill our promise, we will make the situation right.

Event Cancellation or Postponement
SSA reserves exclusive right to modify, postpone/reschedule or cancel programs for any reason, including but not limited to emergency, inclement weather or other 'acts of God'. If there is an event cancellation, every attempt will be made to reschedule, and registration fees will be applied to the rescheduled event date. Any travel, lodging, or incidental expenses incurred related to a cancelled event cannot be refunded under any circumstances. 

Consent to Use of Photographic Images
Registration and attendance at, or participation in, SSA meetings and other activities constitutes an agreement by the registrant to the use and distribution of the registrant or attendees' image or voice in photographs, videotapes, electronic reproductions and audiotapes of such events and activities by SSA. 

If you have any questions, please contact Marie-Louise Rankin, SSA Executive Officer.


Powered by Wild Apricot Membership Software