Great explanation and the visuals take it to the next level. Thank you very much!

I have yet to see such a good video explaining LMM. Thanks from Zürich!

Excellent job explaining this in an understandable way! Thank you so much!

Superbly explained with great visuals

This is the best video I’ve seen on this topic.

Spectacularly well explained. Thanks for that.

Excellent explanation! I wish I had seen this video years ago, I would have saved myself a lot of time to get in to the topic...

Watched four videos on this, and this was the one that made it click. Thanks for your relatable breakdown!

I finally understand the topic! Thank you so much,

Fantastic explanation! Thank you :)

Thank you very much. This is very helpful for a person who has no prior knowledge to statistic. This will definitely help my research project.

Incredible video!!

this is great, thank you.

Great explanation !!! Thank you

Thanks, super clear!

Great video!

I really enjoyed your video and I have a few questions. Could you please explain when a linear mixed model can be used in situations where there are missing values, such as when only two time points are measured and some subjects are not measured at one of the time points? Also, I'm curious if the random intercept model(two measurements) still has the same p-value as ANOVA with paired-t when dealing with missing values. Thank you!

Hello Andreas. First, congratulations on your magnificent videos. They are crystal clear and a very good resource. I have made some calculations and it seems that the linear regression model matches the one you showed at the beginning of the video, although the intercept I calculated is 93.0. The rest is the same as you. I don't know if I am missing something. Thank you!

Excellent explanation! If we have a linear model lm2=weights ~ weeks + personId, then Sum of Squared Error or Residual Standard Error will be 11.8 which is close to LMM model with random intercepts. And Even more if we use a interaction terms "weeks*PersonID" then SSE is 4.5. So, how do we explain the benefits of LMM for these models?

Great video! What if I have independent samples across 3 times measurement time?

Hi. Great video. Are there any slide or notes for these lectures that are available????

interesting example! But I still have a question: in this example, reason for causing failure is that individuals have different weights at begin, if we use traditional liner model and just adjust this factor as a covariate, is that ok? and what's the different between the two models?

Thanks for the explanation, if the subject treat as random effect, then what is the fixed effect?

Can this be used as a replacement to T-tests when the samples are small (e.g., 16 per group)?

Can you recommend a text (in english) that addresses the broader subject of mixed effects models in just as an understandable way as your video?

can I find somewhere examples of random coefficient models where the variable of the random coefficient is not continuous but categorical? ideally written with STATA or SPSS?

Are mixed effect models same as random parameter models?

At 11:26, -6.0, -18.0 and -21.0 are not intercepts. They are slopes of Subjects 2, 3 and 4 with the slope of Subject 1 as the reference. Please correct me if I am wrong.

I can just don't get how you explain the interaction time:group of intervention in a simple clinical trial in a longitudinal study. When is significant time:group of intervention, does it mean that time has an effect on the results? But the patients are under a trial intervention? This means something I guess. How would you explain it?

At 8.21, you say that "multiple leaner regression model does not give an overall intercept". This is confusing as it is actually the expected value of the response variable when all predictors equal zero. Please clarify.

There is absolutely nothing wrong with the fixed effects model. Using a dummy variable for each individual captures the dependence within individual that would otherwise be there. In fact, it is a more robust solution since the normal assumption may be incorrect. In the example you give, there is absolutely no reason to make the almost untestable assumption that the intercepts a a draw from a normal distribution. And if we are are interested in the effect of dieting, why would we make an extra untestable assumption. Bottom line is that RE models are ill-advised in most situations. BTW: I am a Professor of Statistics.

Thank you so much for your excellent explanations. Can you please create a video that explains in simple terms that when we should consider a variable "random" and when "fixed"? As some feedback, is it possible to pronounce "d" in the word "moDel"? You pronounce it "moWel". This and other odd pronounciations distract the listener.