R demo | Paired Wilcoxon Signed Rank Test | How to conduct, visualise and interpret

[@yusmanisleidissotolongo4433]

Hello Professor Yuza, I have to confess that I DO admire you and respect you profoundly.

[@yuzaR-Data-Science]

Thanks 🙏 for the nicest feedback ever!!! Greatly appreciate that and glad you find it useful!

[@Adeyeye_seyison]

Thanks a million sir for ALL you do and represents and your value adding contents and tutorials...

[@yuzaR-Data-Science]

You are very welcome 🙏 thanks for watching

[@Rumil_]

I really like how you explain your thought process, in that why you used shapiros and what happens next if it was or wasnt normal. Great stuff!

[@Rumil_]

Wow I just finished the video and you even mention what would happen if we continued doing the parametric test and what the wrong value would be. Honestly it was truly easy to follow especially with the outstanding editting work. Where can I support and follow more of your work. I find that you explain and visualize things so perfectly that i want to continue learning more from you!

[@yuzaR-Data-Science]

Glad you enjoyed it, Rumil! A 1000 thanks for a nice feedback! Feel free to add any critics also, so, I can improve future videos. Cheers!

[@yuzaR-Data-Science]

Again, 1000 thanks, Rumil. Here ist the place to stay and in the video description is the link to my blog, where you can see more details, R code for copy-pasting and other articles. If you didn't see the Deep Exploratory Analysis video, I could imagine you might enjoy it. The editing is not as fancy, but the material (also in the blog) is very useful in my humble opinion.

[@MySongboy]

Thank you for explaining, I now can use the test much broader.

[@yuzaR-Data-Science]

Glad it was useful! More to come!

[@aminebahmed7526]

Thank you for sharing, It's a joy to watch your content. I noticed that you haven't checked the symmetry of the distribution of diffrences before conducting the test (which is one of the test assumptions), is there any evidence for suitability of not doing so ?

[@yuzaR-Data-Science]

Thanks for such a nice feedback! :) the normality of the difference was checked, right? and it is enough for me enough to decide that I go with non-parametric test. further assumptions are not nessessary to check in my humble opinion

[@JDEGENFELLLNER]

Love your videos!

[@yuzaR-Data-Science]

Thanks, Jürgen! Glad you enjoy it!

[@statlab_stat.solution]

I wish you would have at least 1Biliion subscribers.

[@yuzaR-Data-Science]

Thank you 🙏 soo much! May be one day ;)

[@lidiabezerra32]

Hi, thank you so much for the class. I have a question. How can i invert the category of time? I need to show “before” x “after”, not “after” x “before” (as shown on the vídeo)

[@yuzaR-Data-Science]

Hi Lidia, thanks for the feedback! :) changing order is easy with levels of factors: your_data %>% mutate(time = factor(time, levels = c("before", "after"))

[@TheGoooonrider]

Hello and thank you very much for this video and your article! I want to use the ggstatsplot package for a one-side paired wilcoxon signed rank test. But I do not understand the meaning of the p-value in the plot (in your article: 0.23) in comparison to the p value of the single-sided test (in your article: 0.0114). How do I have to use the ggstatswithin command to obtain the p value of the single-sided paired test? Thank you very much in advance! Greetings Max

[@yuzaR-Data-Science]

Hey Max, this feature was already requested for ggstatsplot, but the author turned it down, since it can't be applied for robust and bayesian methods. Therefore, it's unfortunately not possible with ggstatsplot. I personally don't use them, because the are not as popular and can confuse the reader, and I also don't see them in the medical papers often. So, 0.023 is the two-sided p-value. While the 0.990 and 0.0114 are both one-sided, only in two different directions. Hope that helps! Cheers!

[@isabellafernandez6159]

This is very helpful thanks. How do I remove the statistics at the top of the plot? :)

[@yuzaR-Data-Science]

Thanks! My colleague also sometimes ask to remove it. And you can easily do that (code below), but it's actually the most valuable part. So, if you understand every number there, you can explain it in the legend of the pic. But if you still wanna remove it: ggwithinstats( data = d, x = speed, y = score, type = "nonparametric", results.subtitle = F )

[@jhoandysroyet4667]

Excellent video. I really like your explanation. I am learning to use r and I have a question. How to perform nonparametric analysis of variance in factorial designs?

[@yuzaR-Data-Science]

Great question. There are lots of ways. One particular papers about this in R is here: journal.r-project.org/archive/2016/RJ-2016-027/RJ-2016-027.pdf

[@hikeaway1596]

🙏👍💪😎

[@yuzaR-Data-Science]

thanks!

[@mz555]

the interpretation of p-values is incorrect. they do not express evidence. that is a fisherian interpretation which does not follow from the neyman-Pearson NHST framework.

[@yuzaR-Data-Science]

Thanks for your suggestion! I think I know the reference, but could you tell me the exact reference you quote? I would love to have a look!

[@yuzaR-Data-Science]

Besides, IMO p-values is one of the most confusing and divisive concepts in statistics. There are plenty of definitions of p-values, but not a single correct, widely accepted and accessible definition. Some are more correct but less understandable, some are more intuitive (like my in my videos), but surely oversimplified. Bayesian statisticians run away from p-values. While most of people who use p-values are not statisticians at all (scientists) and they use it in a worst way possible - having this one “magic” threshold of 0.05 and treating p-values as black and white decision & publication making machine. Thus, the ambition to be “correct” on this channel would be unrealistic for me, so, the only goal I have here is to try to be less wrong and to try to bring people one step away from the black and white p-value treatment.

[@yusmanisleidissotolongo4433]

@@yuzaR-Data-Science I agree with you. Besides, it is easier to criticize than making a case by supporting with evidence. Many of of are GRATEFUL at you Pr. Yuza.