Clear, concise explanations. Cannot thank you enough for this. It is rare, in my experience, to get this quality instruction. Looking forward to more videos!

The videos on this channel offer the most interesting and best explanations on these topics, and I've seen many. I''m a 4th year PhD student of behavioral ecology and a 23 year veteran teacher. Take it from me: you are a highly talented instructor! I look forward to more videos! Suggested topics: Bayesian equivalents of frequentist statistical procedures.

A lecture of around 20 minutes is the perfect length!

These videos are great! I'm glad to find an ecology-focussed series on statistics!

Thank you for video. It was amazing to clarify the main differences between the models.

Ur explanations r the best! Thanks a lot

Thank you very much! I've learned a lot.

Great lecture Sir!

Thank you

thank you.

It should be noted that the families can take more links (i.e. you can calculate family=gaussian(link="log"))

Hello, how can you compare different AICs with different dfs ?

I always heard that looking at the base R diagnostic residual plots for generalized linear models isn't useful in the same way it is for general linear models? would like confirmation of the oppisite as it would make my current stats work easier haha

Hi! I would like to know the AICs from different distributions/link functions are comparable in the first place? Doesnt the likelihood function differ? Can you please provide a reference, where it is explained that it is possible to do model selction in that way? Would help me a lot!!! Thank you!

@mjf6125

4 жыл бұрын

It's my understanding that while the likelihood functions are different, they are attempting to solve the same thing. That is 2(log(p(y|saturated model)) - log(p(y|current model))) or something close to that ha. Essentially is saying the difference between the saturated model (the model that EXACTLY fits all data points) and the current model in question. This is the deviance. AIC is deviance but with a penalty on the number of predictors (since increasing the number of predictors will always lead to a decrease in deviance due to overfitting). These AIC then can be compared across models. But again I'm no expert so take everything I said with a grain of salt ha.

Thanks! I would ask when I can use the model like lm, glm.. ? Is it instead of ordinary analysis?

How do you get the residual plot for the first example?

@henrybirt5896

6 жыл бұрын

www.r-bloggers.com/visualising-residuals/

So the evaluation of GLM model is done by comparing AIC values? Do we use R2 or R2 adjusted as well?

I also would like to see how the final plot looks like... you only showed residual plots

@bholly24

5 жыл бұрын

Generally, papers don't actually create final plots for a glm, instead the glm table is presented (especially for more complicated mixed models). However, I have seen GLMs plotted; check Getting Started with R: An Introduction for Biologists.

Hi, why is R studio producing different results even though I am using the same call and data.

Please interpret the result of gamma with log link coefficient results

I wish you could explain more about the AIC number. Where did u get it from? Is it model selection?

Five stars

Audio not good !

@methodsinexperimentalecolo5000

5 жыл бұрын

Just increase the volume

@WahranRai

5 жыл бұрын

Do you think i dont do it before writing the commentary !!!!!!!!!!!!!!!!!!!!!!!!

@jacobm7026

4 жыл бұрын

@@WahranRai did you try increasing the volume?