Thank you for this video. Indeed one of the most challenging parts with working with statistical models in R is that the diagnostics are spread out across multiple packages! keep up the great work.
I enjoy watching your tutorials again and again. Thank you for your amazing tutorials. Just one question, can I use this package to check the assumptions of logistic and Cox regression models?
Thanks, Muhammed! Sure you can for logistic regression. The cox one is more complicated. Important to notice: we are not allowed to put the responsibility on the tool. You have to know what those assumptions are and use the check_model() or other tools as a help, not as a controller. Because not all assumptions for all models are yet implemented, but for most common models for sure. Cox- models less.
Thank you very much for the detailed explanation. I had been using different functions/packages for checking the assumptions, this package made it a lot easier.
Hey Suhail, sorry for the late replay, youtube did not show me your comment. For those models you also use the "anova(model1, model2)" from "stats" and "compare_performance(model1, model2)" from "performance" packages. Cheers
Great question. You actually can't. And you don't really need to. What I do, when assumptions of lm or lmer are not satisfied, I create the lmrob or rlmer model and compare_performance(normal, robust, rank = T) with this function and see, which fit's the data best. You can also go to the alternatives of lm or lmer, like quantile regression, where you relax most of the assumptions. cheers
Thank you for this video. Do you know if there is a function for saving the panel aggregate in a single figure? ggplot2::ggsave only saves the last one. Cheers!
This looks like a great package, however I can't get it working on a linear model I tried. It produces this error message: "Error: `check_model()` not implemented for models of class `lm` yet."
Thank you very much for your useful videos. I have questions concerning performance package, can this package fix or overcome if there are violations of assumptions regression?. If not, what package is the best for that? thank in advance then..
Hey, "performance" only checks assumptions. Based on these checks you decide which model to use. For examples, you have outliers, use robust model, you have non-normality of residuals, use quantile regression. It of coarse depends on the response variable, if numeric, then usually many assumptions are violated. In this case I'd recommend quantile or bootstrapp regression. It's still the same components of the model, but you'll get a better result. If your response is binomial, then there are almost no assumptions. Thanks for watching!
Question: Can we use "package {performance}" for forecasting in Python? If yes, could please share some links/tutorials? If No, why not/ reason? Btw: Thank you so much for this tutorial. Very helpful.
Thanks for the feedback! No, it's a young R package, still in the development by a scientist, not by a programmer. And most stats in science is done by R, not python. So, it is not for python yet. But may be in the future one day. Cheers
![เขามัทรี - เอ็กซ์ ศุภกฤต - จอนนี่มิวสิค [ Official MV ]](http://i.ytimg.com/vi/AS9ufKuQIVs/mqdefault.jpg)
Thank you for this video. Indeed one of the most challenging parts with working with statistical models in R is that the diagnostics are spread out across multiple packages! keep up the great work.
Yes, exactly! By the way, since I made this video, the package evolved and became even better. Thanks for the feedback! Cheers
I really don't know how to thank you. Please, keep making video content of useful packages in R.
Again, I'm very grateful for your work.
😊😊😊😊
More to come! Thanks for watching!
This is exemplary! Made regression a lot easy and enjoyable!
Thanks for nice feedback, Moses! I am very glad it's useful!
Great overview with examples of the R package {performance}. Thank you Yury!
Thanks Andrea! Glad it was useful for you! The package actually developed since the time I made this video, so it's even better now.
❤❤❤ excellent review what a useful package and you explained it all so quickly and concisely I appreciate this
Thank for your nice feedback, Melissa! And thanks for watching!
I enjoy watching your tutorials again and again. Thank you for your amazing tutorials.
Just one question, can I use this package to check the assumptions of logistic and Cox regression models?
Thanks, Muhammed! Sure you can for logistic regression. The cox one is more complicated. Important to notice: we are not allowed to put the responsibility on the tool. You have to know what those assumptions are and use the check_model() or other tools as a help, not as a controller. Because not all assumptions for all models are yet implemented, but for most common models for sure. Cox- models less.
Amazing
Thank you! It's an older one. You might like the newer videos even more ;) Cheers!
Thank you very much for the detailed explanation. I had been using different functions/packages for checking the assumptions, this package made it a lot easier.
It’s my pleasure, Suhail! I am very glad it’s useful!
@@yuzaR-Data-Science Which package can be used to compare proportional odd logit regression model (polr, in package MASS)?
Hey Suhail, sorry for the late replay, youtube did not show me your comment. For those models you also use the "anova(model1, model2)" from "stats" and "compare_performance(model1, model2)" from "performance" packages. Cheers
Fantastic! Thanks.
Glad you liked it! You are welcome!
WOW! AMAZING!
thanks! hope the rest is also useful!
thank you so much!
You're welcome!
Hello yuri, thx for u videos.
You are welcome, Dim!
How can I check the assumptions if I run a robust regression?
Great question. You actually can't. And you don't really need to. What I do, when assumptions of lm or lmer are not satisfied, I create the lmrob or rlmer model and compare_performance(normal, robust, rank = T) with this function and see, which fit's the data best. You can also go to the alternatives of lm or lmer, like quantile regression, where you relax most of the assumptions. cheers
Hi! Wich program are you using to run the package?
R
Thank you for this video. Do you know if there is a function for saving the panel aggregate in a single figure? ggplot2::ggsave only saves the last one. Cheers!
Hey Adrian, the "ggsave" unfortunately doesn't work. don't know any other function which could. So, I click Export and save it manually. Cheers!
This looks like a great package, however I can't get it working on a linear model I tried. It produces this error message: "Error: `check_model()` not implemented for models of class `lm` yet."
It happend to me too, but after I updated all (or most) of the packages, it worked perfectly.
@@yuzaR-Data-Science I'll give that a go. Thanks.
Thank you very much for your useful videos. I have questions concerning performance package, can this package fix or overcome if there are violations of assumptions regression?. If not, what package is the best for that? thank in advance then..
Hey, "performance" only checks assumptions. Based on these checks you decide which model to use. For examples, you have outliers, use robust model, you have non-normality of residuals, use quantile regression. It of coarse depends on the response variable, if numeric, then usually many assumptions are violated. In this case I'd recommend quantile or bootstrapp regression. It's still the same components of the model, but you'll get a better result. If your response is binomial, then there are almost no assumptions. Thanks for watching!
Thank you very much @@yuzaR-Data-Science for come in handy explanation..
@@muhammadhudaya4198 you are welcome!
Question:
Can we use "package {performance}" for forecasting in Python? If yes, could please share some links/tutorials? If No, why not/ reason?
Btw: Thank you so much for this tutorial. Very helpful.
Thanks for the feedback! No, it's a young R package, still in the development by a scientist, not by a programmer. And most stats in science is done by R, not python. So, it is not for python yet. But may be in the future one day. Cheers