What is GEE (Episode 27)

[svalbard]

"you can replace the gaussianity with asymptotics..." Music to my ears. Also, for other noobs such as myself: 7:14, you don't exactly need to "guess" but you should experiment, and using the guidance via QIC given in Cui's (2007) article "QIC program and model selection in GEE analyses" is a great step for informed selection of correlation structure AND model selection (via best model fit). If you don't have Stata you can do the same thing in R using the "geepack" and "MESS" packages. Good luck fellow modelers!

[Rumil_]

Hey awesome video. Quick question, you mentioned in a longitudinal setting where you have small samples with much much more observations within them, you would likely use another method. How does asymptotes facilitate the method of estimation in this case and what would you recommend?

[ericaverde]

Thank you, this gave me a glimmer of hope for my dissertation! 🥳🥳

[tothebee]

Great video, love that you shared this with the whole internet vs just your students :)

[FlowerGiu]

Thank you very much! Great intro to GEE, just the video I was looking for!

[yavarjn]

Could you please cite the book you mentioned here?

[KamarulImran]

Yup. Peter Diggle from Lancaster University. Was my phd supervisor. Anyway, awesome video on gee

[sammyguitar]

Hey Brian, thank you for your nice statistic videos! I'm from germany and writing my master thesis about the dictator game and could need some help. We had one session with 48 participants. They had a questionnaire with 8 questions: How much money would they give to another person with 8 different budgets (-70 -25 -5 0 30 45 65) ? So its like a repeated measures design. I want to test the regression of the different budgets and the amount of money, each subject would give in each situation. Is it correct to use the GEE method in Stata for this kind of regression?
Greets from germany and thank you very much!

[mikkolokolo]

Hello everybody, i watched this video in order to understand the basics behind the GEE, didnt really understand much.
I'm doing a study where i compare infertile patients. So there's a way to treat the infertility with inseminations but sometimes it takes more than one try in order to succeed. there are differents ways these patients can be monitorised. shortly, here the study that introduced me to this theorie, i'd appreciate a lot if someone could explain some of the basics..
thanks in advance :)