sad truth is that I did mixed models once for a publication and one of the reviewers said the statistics section is hard to understand and not common, so i should use anova instead... cheers to the standards of nowadays science edit: After submitting to a journal in another field where I knew from a colleague that the standards in statistics are a little higher, I had no problems anymore.
Good data allows organizations to establish baselines, benchmarks, and goals to keep moving forward. Because data allows you to measure, you will be able to establish baselines, find benchmarks and set performance goals. A baseline is what a certain area looks like before a particular solution is implemented.
You talk about dependence within individuals. Why can you not include a dummy variable for each individual and, if desired, an interaction of this dummy variable with the covariate? This is a FE model. What is the value in pretending that the individual parameters follow a normal distribution (when they might not)?
Hi Chris, the approach you describe only works if "sphericity" is satisfied, for which you need equal variation on the dependent variable for each cluster (each individual in this case). While Mauchly's test tries to identify whether sphericity is violated, a mixed model assigning a random effect to the clustering variable avoids this requirement
Great explanation! I find interesting that in this explanation it may be implied that random effects models (aka multilevel or mixed effects models) may be favoured to fixed effect ones, which instead through a lot of information away. Some researchers especially in econometrics instead would make the distinction between FE and RE models (rather than random and fixed effects) and favour fixed effects
From the econometrics viewpoint, the main issue and reason of unpopularity of RE in the discipline is a very strong assumption of zero covariance between individual- and group-level variables. Highly unrealistic in the wild, highly important for consistency of the estimator.
Firstly, i would like to thanks for you interesting study, my data have two land uses(exclosure and non exclosure) with three site in each land use how to arrange my data and make analysis using liner mixed effect model
Hi Sir, if Hausman test indicates that fixed model is more appropriate than random effect model, and if in that case, in data time period (T) > cross section units (N), which FEM is to be chosen: time (T) FEM or Cross section (N) FEM?
Hey! thank you so much for this explanation it was truly helpful. I was wondering if you could answer a question I had about the topic. What if you wrongly assume a factor to be of random effect how would that affect your results if at all?
That is why we have many independent variables to capture the random effect.. but what i was expecting how these fixed vs random effecting impacting the model.. where we already tried using many independent variables
Thank you for the explanation, this video was very easy to understand!
I love you Tom, you managed to explain this incredibly important point to me in such an eloquent manner that I finally understand its significance!
love him too
I loved the video. Thank you Tom!
Such a clear explanation! Very helpful.
Great Video! Please upload the Mixed Effects one
sad truth is that I did mixed models once for a publication and one of the reviewers said the statistics section is hard to understand and not common, so i should use anova instead... cheers to the standards of nowadays science
edit: After submitting to a journal in another field where I knew from a colleague that the standards in statistics are a little higher, I had no problems anymore.
Its a major limitation of the peer review process...
Amazing explanation! I wonder if the video about mixed models is already out? I could not find it under the youtube page of Univ. of Nottingham...
We found these other videos with Tom Reader ruclips.net/video/z45LUip6RcI/видео.html and ruclips.net/video/PyNzbDbjs1Y/видео.html if they help at all.
This was fabulous! I really enjoy your style of presenting. It is clear, challenging, and well-crafted.
Hello!! can one use a fixed effect regression on a cross-sectional dataset, if yes how?
Good data allows organizations to establish baselines, benchmarks, and goals to keep moving forward. Because data allows you to measure, you will be able to establish baselines, find benchmarks and set performance goals. A baseline is what a certain area looks like before a particular solution is implemented.
Big up the top g Tom, shelling stats like it's Mario Kart. GG
Hi, I'm From Comoros. Thanks for the video, it was crystal clear !!!
This is brilliantly done. Wonderful presentation!
Great video! Well explained, thank you. I wonder if at 6:00 it is going about the random effects and not bias measurement? Thanks!
No! We need the mixed effect model video. This is the clearest explanation I've heard.
Sir, please give the lectures in written form also
This is really well done! Great job Tom Reader!
Excellent explanation of effects in statistical models! Huge thanks Tom, you are the best!
You talk about dependence within individuals. Why can you not include a dummy variable for each individual and, if desired, an interaction of this dummy variable with the covariate? This is a FE model. What is the value in pretending that the individual parameters follow a normal distribution (when they might not)?
Hi Chris, the approach you describe only works if "sphericity" is satisfied, for which you need equal variation on the dependent variable for each cluster (each individual in this case). While Mauchly's test tries to identify whether sphericity is violated, a mixed model assigning a random effect to the clustering variable avoids this requirement
Great explanation! I find interesting that in this explanation it may be implied that random effects models (aka multilevel or mixed effects models) may be favoured to fixed effect ones, which instead through a lot of information away. Some researchers especially in econometrics instead would make the distinction between FE and RE models (rather than random and fixed effects) and favour fixed effects
From the econometrics viewpoint, the main issue and reason of unpopularity of RE in the discipline is a very strong assumption of zero covariance between individual- and group-level variables. Highly unrealistic in the wild, highly important for consistency of the estimator.
Really clear explanation! Thank you!
Firstly, i would like to thanks for you interesting study, my data have two land uses(exclosure and non exclosure) with three site in each land use how to arrange my data and make analysis using liner mixed effect model
Hi Sir, if Hausman test indicates that fixed model is more appropriate than random effect model, and if in that case, in data time period (T) > cross section units (N), which FEM is to be chosen: time (T) FEM or Cross section (N) FEM?
Very clearly explained, cheers
Hey! thank you so much for this explanation it was truly helpful. I was wondering if you could answer a question I had about the topic. What if you wrongly assume a factor to be of random effect how would that affect your results if at all?
That is why we have many independent variables to capture the random effect.. but what i was expecting how these fixed vs random effecting impacting the model.. where we already tried using many independent variables
Great Lectures. Many thanks. Is there a sequel into explaining more about Mixed models
can "nurse" be treated as a random effects if there are only 2 nurses?
Hope there was a link to the next video for the mixed model
Extremely good video, Mr. Reader. Thank you so much.
Thank you so much for simplyfing such topic.
Thank you for this clear explanation!
Very clear explanation . Thankyou !
Thx Tom, great explaination :) and well pronounced btw!
What an awesome video! Thank you!
Thanks Tom. Great explanation
Such a great explanation and I finally understood this importing thing
Great theoretical background
What an excelent video, thank you very much
Thank you very much, sir
Excellent video and crystal clear explanation
Fantastic! Thank you!
very well explained
great presentation!
That was very clear. Thanks a lot!
Thank you so much!
very good
Thank you, sir
Thank you sir
great
Superb , lucid presentation on an all too often neglected topic in stats.