22:28, looking at your model, I don't see any control variables. Aren't we supposed to control for extraneous factors such as the subjects' age or gender?
Thank you very much for these amazing informative videos! The summary() function outputs comparisons of groups C5-8 and paraplegia with the reference group C1-4 in terms of time group interaction. How could one compare C5-8 to paraplegia in terms of interaction, both linear and quadratic? Thank you!
Hey! Amazing video, this really helped me with a previous assignment. Just have a question whether you think using ANOVA would be appropriate to test for whether a plant’s “health” (categorical data, leaves are either: green, grey, yellow/orange) with the total number of individuals in a population?
Thanks! Yes, if you have a one categorical factor of Health (Green, Grey, or Yellow/Orange) as a predictor and Total Number as your outcome, then factorial ANOVA certainly could be appropriate. One thing to check is how the "count" variable is distributed. Counts can be very skewed which might lead to a violation of normality for your residuals. ANOVA is fairly robust to violations of the normality assumption, but still a good thing to check. : )
Hey. I usually don't leave comments here, but I just wanted to say you having so little views is a pity as after scrolling through dozens of videos on this to no avail, you explained it perfectly and I was actually able to finally understand. So I guess don't give up? <3
I believe if the time variable is not the variable of study, we don't care that much about multicollinearity or we should at least prove that mean centering improves the VIF in order to use it. If time variable is merely a control variable, well....we don't care about multicollinearity.
Thank you for these wonderful videos, they are very helpful. Just a small comment: it would be very helpful to discuss in which situation which model should be used. For example, when the researchers should use a nonlinear negative exponential mixed effect models and how they should choose random and fixed effects. Unfortunately, I did not have time to go through all the videos. Please skip my comment if you have already discussed these points.
Thank you for a series of very thorough and clear videos. I have been wrestling a dataset of my own and I found that working along with these videos, yet with my own data, really helped. Do you have any words on using the 'generalized variance inflation factor' or GVIF in order to assess multicollinearity in this dataset?
I am glad they were useful. I am not very familiar with GVIF (as opposed to "classic" VIF) but there is a helpful discussion where John Fox himself chimes in here: stats.stackexchange.com/questions/70679/which-variance-inflation-factor-should-i-be-using-textgvif-or-textgvif More generally, I would say this is going to be a particular concern in models with interactions or polynomial terms, so I would definitely recommend contrast coding your factors (as opposed to treatment coding) and mean-centering continuous variables as a preventive step to reduce variance inflation in those models.
This is a really informative and helpful video! I wondered if you could point me in the direction of resources to support an a priori power analysis for mixed effects models for longitudinal data?
Hi Gemma, that can quickly become a complicated subject! In brief though, GLIMMPSE is a web based platform that is free and there are since free packages in R (like “simR”) … and then PASS is a paid program for a lot of power calculations (including mixed models) that makes it easier but only covers limited models. I will try to do a video doing a power analysis through simulations in R soon.
Hi, thank you for your video. Could you make a video of linear mixed effect with repeated measures, like in crossover designs? I really will appreciate it.
Thanks Nathaly. Please see my other video of mixed-effect models for factorial designs: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-x467LStTtHU.html You might also be interested in this pre-print (not peer reviewed yet) on arXiv: arxiv.org/abs/2209.14349