Dear Dr. Zablotski. Thanks a million for another phenomenal video. I would love to listen to your suggestions on dealing with p values while working with large samples, in particular on adjusted p values. Much love and respect.
Thanks again Sadat! Greatly appreciate your feedback! I plan to do a p.values video, but until then, the p.values in large samples are often significant, so you could reduce the threshold for significance, from 0.05, to 0.01 or lower. And I always use Banjamini & Hochberg instead of Bonferroni adjustment. Hope that helps! Cheers
I stumbled upon the ggbetweenstats function a few months ago but didn't appreciate all of its nuances, which are so nicely laid out in your video. Thanks!
Wonderful! As mentioned in the previous comment, the newest videos have even more details and explanations of every number, including Bayesian test below the picture, effect-size etc. Check them out if you wish. Thanks again for watching!
Hi Yury, your videos are very helpful. Why do your confidence intervals go to "[0, 0]" at the end of the video? I found this was happening with my data as well, and I'm getting p-values of "1"
Thanks for the feedback, showtyme11111! The 0, 0 CIs are for the effect size, which is also zero - means - not there. That's totally legit, especially when the p-value is high, like in your case - 1. It simply means - there is absolutely no difference in a sense of stat-significance and no effect (which per se is the difference). P-values can be significant, but they don't tell anything about the effect size. Hope that helps. Cheers!
I guess there are no significant differences between the groups then. There is an argument you can use to show all the comparisons: pairwise.display = "all"
you are welcome! I think you can change the appearence of the plot with normal ggplot2 commands. I only used a different theme somethimes, but never tried to enlarge the borders. It might be possible though. You can also have a look at the ggstatsplot arguments list, there are a lot of things you can adjust. Just google it and you'll find the github page of ggstatsplot package, then progress to the command you are interested in and you'll see all the options
@@yuzaR-Data-Science Yes, of course. You can enlarge the violin borders to make them more visible by increasing the size in the code: violin.args = list(width = , alpha = , size = ). For example: ggbetweenstats( data = data, x = variable_x, y = variable_y, type = "np", violin.args = list(width = 0.4, alpha = 0, size = 1),
great explanation. I intalled the ggstatsplot package but unable to use any of the functions. R always sent and error meassage saying " function not found" even after calling both ggplot and ggstatsplot. Can yu suggest a solutuion?
thanks, it definitely works since I use it everyday. so, first, update everything: R, Rstudio, all packages. Secondly there might be some dependencies. Thus, google the function which does not run and find out which package is it from, install this package. ggstatsplot became even better since the time I made a video, so it's worth your time to bring it to live, and again, your problem is purely technical, not from ggstatsplot. cheers
Dear Dr. Zablotski. Thank you so much for your explanation! I have a question, in the package ggstatsplot do we have some function similar to facet_wrap of ggplot?
Yes, absolutely. It has "grouped_ggbetweenstats" and similar function, all starting with "grouped_*". Here is an example: indrajeetpatil.github.io/ggstatsplot/articles/web_only/ggbetweenstats.html
@@yuzaR-Data-Science Thank you a lot for your reply! So, for example, I want to see two different types of lung diseases according to the expression, related to different marks. So I can write: grouped_ggbetweenstats(data = TB_pleural, x=Diagnosis, y= Expression, grouping.var = Markers), in theory, this will give a series of plots related to the markers, right?
@@yuzaR-Data-Science yes! I think the code is right. But after running appears a error saying: "The Rstudio 'Plots' window may be too small to show this patchwork. Please make the window larger." I already did the window larger. But the same error continues to appear!
Dear Sir, I'm working on my data, obtained through Likert scale of 5. I want to conduct Mann Whitney test through R. I searched a lot but could not find anything. Please any guidance or reference, as I have no experience of R software. Thanks in advance
Hey Zeb, just use the code I provided in the video. The package was updated and now, if you use the argument, type = "np", you'll get the Mann Whitney test. Wilcoxon and MannWhitney are actually synonyms, check wikipedia please. So, you are good to go with "ggstatsplot" package! Hope that helps! Cheers, Yury
HI! I know this is an old video by now, but I want to remove the red dot of the mean, but when I use mean.plotting it won't remove, has the code for this been changed?
How many observations you have depends on your table and it's form, ggbetweenstats does not change the data, it takes it as it is. Or you might have NAs.
I think it's ok then, in the code before ggbetweenstats, you might have filtered observations. I can't say anything without seen the data. But fften a simple misspelling can be a problem. Try another small dataset (like mtcars or similar), where you know, that you have groups in one column, and values in the other.
Dear Dr. Zablotski. Thanks a million for another phenomenal video. I would love to listen to your suggestions on dealing with p values while working with large samples, in particular on adjusted p values. Much love and respect.
Thanks again Sadat! Greatly appreciate your feedback! I plan to do a p.values video, but until then, the p.values in large samples are often significant, so you could reduce the threshold for significance, from 0.05, to 0.01 or lower. And I always use Banjamini & Hochberg instead of Bonferroni adjustment. Hope that helps! Cheers
I stumbled upon the ggbetweenstats function a few months ago but didn't appreciate all of its nuances, which are so nicely laid out in your video. Thanks!
Wonderful! As mentioned in the previous comment, the newest videos have even more details and explanations of every number, including Bayesian test below the picture, effect-size etc. Check them out if you wish. Thanks again for watching!
Great explanation, Sir.
Thanks Yan! I am glad it is useful! If you liked the Demo, you might also enjoy this previous video: Box-Plots in R, where I explained even more.
Hi Yury, your videos are very helpful. Why do your confidence intervals go to "[0, 0]" at the end of the video? I found this was happening with my data as well, and I'm getting p-values of "1"
Thanks for the feedback, showtyme11111! The 0, 0 CIs are for the effect size, which is also zero - means - not there. That's totally legit, especially when the p-value is high, like in your case - 1. It simply means - there is absolutely no difference in a sense of stat-significance and no effect (which per se is the difference). P-values can be significant, but they don't tell anything about the effect size. Hope that helps. Cheers!
Good one
Old thought, newer videos might be more useful
thanks
You are welcome 🙏 thanks for watching
Great presentation. When I use the ggstatplot, the comparison between the different groups does not show on the graph. Any adjustment needed?
Thanks
I guess there are no significant differences between the groups then. There is an argument you can use to show all the comparisons: pairwise.display = "all"
Thanks for this educating piece.
Is there a way to thicken the borders of the box plot and violin?
you are welcome! I think you can change the appearence of the plot with normal ggplot2 commands. I only used a different theme somethimes, but never tried to enlarge the borders. It might be possible though. You can also have a look at the ggstatsplot arguments list, there are a lot of things you can adjust. Just google it and you'll find the github page of ggstatsplot package, then progress to the command you are interested in and you'll see all the options
@@yuzaR-Data-Science Thanks a lot. I tried it. I was able to enlarge the violin's borders but not the box plot ones.
@@ibrahimlawan9663 cool, how did you do it? could you comment the code? That would be useful for the whole R community here. Thanks in advance!
@@yuzaR-Data-Science Yes, of course. You can enlarge the violin borders to make them more visible by increasing the size in the code: violin.args = list(width = , alpha = , size = ). For example:
ggbetweenstats(
data = data,
x = variable_x,
y = variable_y,
type = "np",
violin.args = list(width = 0.4, alpha = 0, size = 1),
@@ibrahimlawan9663 cool! highly appreciate that, Ibrahim!
great explanation. I intalled the ggstatsplot package but unable to use any of the functions. R always sent and error meassage saying " function not found" even after calling both ggplot and ggstatsplot. Can yu suggest a solutuion?
thanks, it definitely works since I use it everyday. so, first, update everything: R, Rstudio, all packages. Secondly there might be some dependencies. Thus, google the function which does not run and find out which package is it from, install this package. ggstatsplot became even better since the time I made a video, so it's worth your time to bring it to live, and again, your problem is purely technical, not from ggstatsplot. cheers
Dear Dr. Zablotski. Thank you so much for your explanation!
I have a question, in the package ggstatsplot do we have some function similar to facet_wrap of ggplot?
Yes, absolutely. It has "grouped_ggbetweenstats" and similar function, all starting with "grouped_*". Here is an example: indrajeetpatil.github.io/ggstatsplot/articles/web_only/ggbetweenstats.html
@@yuzaR-Data-Science Thank you a lot for your reply!
So, for example, I want to see two different types of lung diseases according to the expression, related to different marks. So I can write: grouped_ggbetweenstats(data = TB_pleural, x=Diagnosis, y= Expression, grouping.var = Markers), in theory, this will give a series of plots related to the markers, right?
Yes, that should work. Did you tried it out already?
@@yuzaR-Data-Science yes! I think the code is right. But after running appears a error saying: "The Rstudio 'Plots' window may be too small to show this patchwork. Please make the window larger."
I already did the window larger. But the same error continues to appear!
@@matheusalmeida773 :) no, it's your RStudio's window which need to be larger
Dear Sir, I'm working on my data, obtained through Likert scale of 5. I want to conduct Mann Whitney test through R. I searched a lot but could not find anything. Please any guidance or reference, as I have no experience of R software. Thanks in advance
Hey Zeb, just use the code I provided in the video. The package was updated and now, if you use the argument, type = "np", you'll get the Mann Whitney test. Wilcoxon and MannWhitney are actually synonyms, check wikipedia please. So, you are good to go with "ggstatsplot" package! Hope that helps! Cheers, Yury
@@yuzaR-Data-Science Thanks a lot sir for your kind reply. I will try it. Best wishes
HI! I know this is an old video by now, but I want to remove the red dot of the mean, but when I use mean.plotting it won't remove, has the code for this been changed?
Aah it's centrality.plotting now
ggstatsplot::ggbetweenstats(mtcars, cyl, mpg, centrality.plotting = F)
exactly :)
ggstatsplot::ggbetweenstats(mtcars, cyl, mpg, centrality.plotting = F)
@@yuzaR-Data-ScienceDo you have an updated code for adding the notch? I tried to do it but I think this part of the code is changed too :)
@@akanequeen yes, the notch is gone. you might get it quickly from the boxplot in ggplot2, but I think by now you already have seen these video ;)
i have done everything as you said but its showing me only 30 observations
How many observations you have depends on your table and it's form, ggbetweenstats does not change the data, it takes it as it is. Or you might have NAs.
@@yuzaR-Data-Science its jus showing my rows as 30, for "3. Some College" only. Also the box plot formed is only for "3. Some College" .
@@yuzaR-Data-Science Also, when the violin plot is plotted its showing me for only "3. some college".
I think it's ok then, in the code before ggbetweenstats, you might have filtered observations. I can't say anything without seen the data. But fften a simple misspelling can be a problem. Try another small dataset (like mtcars or similar), where you know, that you have groups in one column, and values in the other.
@@yuzaR-Data-Science Sir, The code
set. Seed(3)
xy%
rename(salary=wage) %>%
group_by(education) %>%
sample_n(30) %>%
filter(education %in% c ("1. < HS Grad ", "3. Some College","5. Advanced Degree "))