THANK YOU SO MUCH! I literally spent five hours reading articles and statistics pages that did no more than confuse me and watching your video was so nice and relaxing for me because not only were you clear but you were able to explain it with a useful example
I don't know why I was having so much trouble understanding this test with written direction but your video was very helpful to me. It presented the necessary info concisely and in a manner east to understand.
Thank you very much for your videos! By writing down the statistical tests it is easier to understand and you explain it very well, slowly and in a very clear way !
Thank you so much for posting these videos. I keep running into explanations with shortcuts because calculators usually do the work now, but I get stuck on the details. Wish I had you as a teacher in high school!
thanks Sir for explaining the topic in such a good manner 👍👍......it is in my MBA syllabus and I am thinking of learning in advance about this topic and after watching your video I can surely say I learned about the Mann Whitney U Test and it will be very helpful for me now to understand when college starts explaining about this topic
Amazing!! I am studying for the 6-sigma black belt exams and this made me understand the Mann Whitney U test. I have subscribed as your explanations even for the Kruskal-Wallis test was equally very clear!
Once again, thanks for the assist : ) This was an awesome and useful video for me. I am great with the parametric statistical methods, but one of the college students that I tutor had questions about non-parametric students. Great explanation!
Thank you so much! I'm taking an online class, and they never mentioned how to do the ranking, so I've always been really confused at how the answer was gotten.
Great explanation! my module provided an Excel spread sheet to do this, on which it said they expect us to also do it by hand but they didn't explained how. Thanks for the help!
At 8:30, he says that because U stat value is lower than U crit value for this particular test, there is a difference in the treatment ranking. In other words, the different scores are not up to chance and they occurred because of the different effectiveness of each treatment. This shows that there IS a difference between the two treatments which we established to be our research hypothesis. (H1) at the beginning of the video. Hope that cleared things up. :)
@@skrimp_ Here in Mann Witney U test and other non-parametric test having a different criteria of rejection/acceptance the null hypothesis because when we use T-test, z-test, F-test and all other parametric test we reject/accept the Null hypothesis on the basis of when critical value is greater then any value(eg:- t,z,F ratio) we accept the null hypothesis and else reject..
@@faiz_fictions I am sorry bro, I am having trouble understanding what you said. From what I understand from this video, the null hypothesis is only correct when U stat is greater than U crit. I could be wrong. Sorry I am not of much help.
Thank you for that immensely helpful video. I have a small clarification question. Let's say there are three scores that tie, does the procedure of calculating the average apply as well? That is, if you had three 2's among your scores, would each get rank 2 by (1+2+3)/3=2?
this video is excellent, thanks! I have a small sample size and R couldn't deal with all the tied values, was able to do it myself in 15 mins following this.
Two questions - 1. Why did you take U statistic instead of Z for comparing calculated and critical values ? 2. U statistic is calculated with the formulae : U1 = n1n2 + n1(n1+1)/2 - Sigma R U2 = n1n2 + n2(n2+1)/2 - Sigma R your formula for U-statistic is different. I could not find that in any statistic book or internet. All of them give the listed above formula. would appreciate if you could answer that. Thanks, Bhaskar
The formula is different, but the values calculated for U1 and U2 are just exchanged. This is because the above formula just subtracts the formula used in the video from n1n2. In the end while selecting the min(U1,U2) it wouldn’t matter if the values of U1 and U2 are exchanged. But still, I wonder why change the formula anyway.
Mr. Eugene if you are a class teacher and you conduct exams for two different classes but in the subject. mathematics two data is different in number one class is 32 and one class is 14. they are both skewed. con you please use this approach to compare the performances between theses classes?
The explanation was very clear, thank you for sharing it. I would like to make a suggestion - include a slightly more elaborate example that connects these numbers to the real world. In this case you mentioned that the numbers came from the results of two distinct treatments. For me it would be more helpful if that was accompanied by something like "The numbers you see correspond to the number of lesions that remained on a person's body after treatment was applied; thus we want to compare the lesion-healing effects of drug A and drug B". This would make the problem less abstract. Otherwise there is a chance that some people will be able to do the number-crunching, but struggle to come up with an answer when asked "so where can you apply this?".
Hi Trisha, The result is decided based on U stat being less than U crit. This means we reject the Null Hypothesis (no difference) in favour of the alternative hypothesis (there is a difference). In statistics tests the convention is that you either reject or fail to reject the Null, we do not say we reject the Alt Hypothesis. Dr E.
Why is the U value being calculated differently in this example? The original paper by Mann-Whitney 1947 uses a different formula. You are also not the only professor I have seen using a different formula.
Thanks :) but I also want to know how to find p-value for mann whitney. I think to find P-value first we need to find z and then by using calculated value of z we can find p-value either by using Excel or table. What is the exact formula to find z.
Hi Orlan, Strictly speaking, you should not use the Mann-Whitney test if there are ties (1, 1, 1 in your case), as the test is not reliable. However, most statisticians are OK to use this test if there are just a small number of ties. Hope this helps, Dr E.
I have a doubt here. When our table value is less than our calculated value, we reject the null hypothesis and vive versa. But in the video, table value is 5 and calculated value is less than 5 so why are we rejecting the null hypothesis
Hi Rupali, We compare the U crit value (from the table) with the calculate U stat. If the U stat is less than or equal to the U crit, the Null Hypothesis is rejected (as in the video). If the U stat is greater than the U crit, we fail to reject Null Hypothesis. Hope this helps, Dr E.
how is that...when stat value less than critical value and we reject null hypothesis.... in other test we fail to reject H0 in such cases....or this is the only case..??
According to Wikipedia Ua+Ub should equal Na*Nb. in my data set i have 2 groups for a total of 126 responses. Na=67 Nb=59. Responses range between 0 and 10 in their value. Group A has: 17 responses of 0 ranked at 11.5 3 responses of 1 ranked at 26 4 responses of 2 ranked at 33 5 respoonses of 3 ranked at 44 6 responses of 4 ranked at 55.5 20 responses of 5 ranked at 78.5 2 responses of 6 ranked at 99 3 responses of 7 ranked at 103 2 responses of 8 ranked at 110 0 responses of 9 and 5 responses of 10 ranked at 120.5 Group B has 5 responses of 0 ranked at 11.5 4 responses of 1 ranked at 26 3 responses of 2 ranked at 33 10 responses of 3 ranked at 44 2 responses of 4 ranked at 55.5 18 responses of 5 ranked at 78.5 1 responses of 6 ranked at 99 2 responses of 7 ranked at 103 7 responses of 8 ranked at 110 0 responses of 9 and 7 responses of 10 ranked at 120.5 So I calculated my Ua and Ub and tried to check whether I did it right, However, when i add Ua and Ub i get 3985, whilst Na*Nb amounts to 3953. So i must be doing something wrong. I have no clue what. Does anyone have any clue what is going wrong?
Happens almost fail-proof. You calculate something wrong, and you know it is but not why... as soon as you start panicking and flailing around until you give up and ask others for help you see your mistake...., Mine was that i calculated the U's wrong. I did Ua= Ra-((Na(Nb+1)/2) instead of Ua=Ra-((Na(Na+1)/2)..... Aaaanyway thanks a bunch for the vid, also helped me alot. Cheers from the Netherlands!
![Mann-Whitney U Test [Simply explained]](http://i.ytimg.com/vi/LcxB56PzylA/mqdefault.jpg)
![Mann-Whitney U Test [Simply explained]](/img/tr.png)
This is the day before my exam and let's just say you turned a 1hr class into 9 min! Thank You!!!
Day of exam💪💪
@@VaporyBeathow did it go
The best video I have found regarding the Mann Whitney U test! Thank you!
Once again, clear, right to the point, easy to understand, only 10 minutes. Thank you for your work!
I'm doing a Thesis for the Dentistry Degree in Italy and you saved me! God Bless You, man!
THANK YOU SO MUCH! I literally spent five hours reading articles and statistics pages that did no more than confuse me and watching your video was so nice and relaxing for me because not only were you clear but you were able to explain it with a useful example
You're a legend. I am embarrassed to say how long I spent trying to figure this out before i watched this video. Thank you.
Thank you so much! My exam is tomorrow and I just spent all day stressing over this and I am so glad I found this video!
My 1 hour tutorial that had no favorable outcome had been solved in 9 minutes. Thank you, great job.
I'm currently writing my dissertation and this was immensely helpful - THANK YOU!
U just turned a 2 and a half hour lesson into 9 mins and I learnt way faster thank you
I don't know why I was having so much trouble understanding this test with written direction but your video was very helpful to me. It presented the necessary info concisely and in a manner east to understand.
Thank you very much for your videos! By writing down the statistical tests it is easier to understand and you explain it very well, slowly and in a very clear way !
I don't have to perform it by hand but your explanation helped me understand what I'm "clicking" in Statistica. That was so helpful, thank you.
Thank you so much for posting these videos. I keep running into explanations with shortcuts because calculators usually do the work now, but I get stuck on the details. Wish I had you as a teacher in high school!
Mr. O'Loughlin, you are my hero. Thank you so much!!
Thank you sir , this video has cleared all my doubts which my university teachers couldn't solved!
thanks Sir for explaining the topic in such a good manner 👍👍......it is in my MBA syllabus and I am thinking of learning in advance about this topic and after watching your video I can surely say I learned about the Mann Whitney U Test and it will be very helpful for me now to understand when college starts explaining about this topic
This man is out here saving lives.
The best video on UA-cam for mann Whitney u test
Amazing!! I am studying for the 6-sigma black belt exams and this made me understand the Mann Whitney U test. I have subscribed as your explanations even for the Kruskal-Wallis test was equally very clear!
You just saved my classes bacon on their psychology coursework, Thank you!
Thanx proffessor u save my life today bcz tomorrow I have exam but bcz of this video everything is fine now thank u so much😍
I'm taking Stats online and I only have my book to guide me step by step and this video helped so much! Thank you!
Well done Eugene on another great video. Clear and concise as always. Please keep up the good work
thank you so much for this. your video series is so clear and concise. you are a legend sir
Once again, thanks for the assist : ) This was an awesome and useful video for me. I am great with the parametric statistical methods, but one of the college students that I tutor had questions about non-parametric students. Great explanation!
Thank you so much! I'm taking an online class, and they never mentioned how to do the ranking, so I've always been really confused at how the answer was gotten.
Thank u for being a genuine life saver...like I can't even express how much you have just helped me THANKYOU!!!!
Thanku so much! Tomorrow is my exam and your video is so helpful ❤
Thank you!!!! This video made everything sound so simple!! Saved my grade!!
You're the best sir. Take a bow.
Amazing video. It's essentialy one of my statistic lectures in a fraction of the time.
Keep up the good work!
Excellent work sir. Understood it in a single watch.
Great explanation! my module provided an Excel spread sheet to do this, on which it said they expect us to also do it by hand but they didn't explained how. Thanks for the help!
Hi! If there are more than two similar ratings in my data, how to rank them? Thank you
Thanks a ton. Please keep making videos on stats and enlighten us through knowledge.
This was extremely helpful! Very clear, simple, easy to understand explanation.
Why we reject the null hypothesis ? when all criteria are favour to the null hypothesis
At 8:30, he says that because U stat value is lower than U crit value for this particular test, there is a difference in the treatment ranking. In other words, the different scores are not up to chance and they occurred because of the different effectiveness of each treatment. This shows that there IS a difference between the two treatments which we established to be our research hypothesis. (H1) at the beginning of the video. Hope that cleared things up. :)
@@skrimp_ Here in Mann Witney U test and other non-parametric test having a different criteria of rejection/acceptance the null hypothesis because when we use T-test, z-test, F-test and all other parametric test we reject/accept the Null hypothesis on the basis of when critical value is greater then any value(eg:- t,z,F ratio) we accept the null hypothesis and else reject..
@@faiz_fictions I am sorry bro, I am having trouble understanding what you said. From what I understand from this video, the null hypothesis is only correct when U stat is greater than U crit. I could be wrong. Sorry I am not of much help.
You are amazing! Best video for exams EVER! Thank you so so much!!!
Thank you for that immensely helpful video. I have a small clarification question. Let's say there are three scores that tie, does the procedure of calculating the average apply as well? That is, if you had three 2's among your scores, would each get rank 2 by (1+2+3)/3=2?
Yes. 😊
Hi, please can you do a video on Wilcoxon signed rank test?
But I think the h stastic falls in the acceptance region when represented in a skewness how comes we're rejecting the null hypothesis
Thanks God I met your video! You're an ANGEL!!!
Great Sir, very simply you taught it, I felt like I learned it in a class.
I was confused about the last bit, thanks for clearing!
Great video, everything is explained simple and steb by step. Thank you!
this was super helpful!! thank you for taking your time to do this im so grateful!!!
You helped me to pass my exam .
Thanks a lot !
thanks so much my exam was is today and i missed the lectures on this saved me
this video is excellent, thanks! I have a small sample size and R couldn't deal with all the tied values, was able to do it myself in 15 mins following this.
Very simplified and well explained. Thank you
So, how many of you are on the day of exam or a day from exam?
Whoa I actually understand. Great explanation sir!
Thank you for your very clear and helpful explanation.
But for a two tailed test , you are supposed to use the higher u calculated value, right??!! so is this a slight error ?
Beautiful, instructive example sir!
Watching this 1hr before exam :). Thx
wow, i wish my stats teacher was as good at teaching as you are
Thank you so much for your great explanation.Thank you for sharing
I am finding your videos very helpful! Thank you for sharing your knowledge!
I performed Mann Whitney U test and achieved statistical significance however my median score is the same. How do I intepret and explain this?
Two questions -
1. Why did you take U statistic instead of Z for comparing calculated and critical values ?
2. U statistic is calculated with the formulae :
U1 = n1n2 + n1(n1+1)/2 - Sigma R
U2 = n1n2 + n2(n2+1)/2 - Sigma R
your formula for U-statistic is different. I could not find that in any statistic book or internet. All of them give the listed above formula.
would appreciate if you could answer that.
Thanks,
Bhaskar
Cause this just would make sense with a sample size bigger than the one in the video.
You use the z statistic if your sample size is at least bigger than 20, in this case in the video, it is not, so he took the u statistic instead.
same question. our professor used that formula so i got confused.
The formula is different, but the values calculated for U1 and U2 are just exchanged. This is because the above formula just subtracts the formula used in the video from n1n2. In the end while selecting the min(U1,U2) it wouldn’t matter if the values of U1 and U2 are exchanged. But still, I wonder why change the formula anyway.
Also because a z statistic assumes a normal distribution but here the data isn't.
I've got an exam after 4 days and thanks your video
Mr. Eugene if you are a class teacher and you conduct exams for two different classes but in the subject. mathematics two data is different in number one class is 32 and one class is 14. they are both skewed. con you please use this approach to compare the performances between theses classes?
Thank you very much! It was a big help for me. You save my assignment.God bless.
The explanation was very clear, thank you for sharing it.
I would like to make a suggestion - include a slightly more elaborate example that connects these numbers to the real world. In this case you mentioned that the numbers came from the results of two distinct treatments. For me it would be more helpful if that was accompanied by something like "The numbers you see correspond to the number of lesions that remained on a person's body after treatment was applied; thus we want to compare the lesion-healing effects of drug A and drug B".
This would make the problem less abstract. Otherwise there is a chance that some people will be able to do the number-crunching, but struggle to come up with an answer when asked "so where can you apply this?".
Love your videos, please keep making them. I have learned loads =)
Thanks so much, you just saved my psychology group´s life!
One correction whenever your stat value is less than critical value then we accept null hypothesis and reject alternative hypothesis
What do I do if my u value is 220.5? I cant find any chart high enough to find the u critical
can you please share about wilcoxon matched pair test? thank you for your attention
Why is it Reject null hypothesis? the alternative hypothesis should be the one rejected..
Hi Trisha,
The result is decided based on U stat being less than U crit. This means we reject the Null Hypothesis (no difference) in favour of the alternative hypothesis (there is a difference).
In statistics tests the convention is that you either reject or fail to reject the Null, we do not say we reject the Alt Hypothesis.
Dr E.
Why is the U value being calculated differently in this example? The original paper by Mann-Whitney 1947 uses a different formula. You are also not the only professor I have seen using a different formula.
You, Sir, are too the point !! Kudos!
Very concisely and clearly explained, thank you
Very clear and strait to the point thanks
Sir thanku for such a comprehensive vedio. Beautifully presented sir
Thank you. I understand this so much better now
Why my table isnt the same as yours? The values at 6 and 6 is not 5 but 29 on mine
Thanks :) but I also want to know how to find p-value for mann whitney. I think to find P-value first we need to find z and then by using calculated value of z we can find p-value either by using Excel or table. What is the exact formula to find z.
what would happen if there are 3 same numbers? for example i have a data for my experimental, we have 1, 1, 1 and for the controlled 3,5,6
Hi Orlan,
Strictly speaking, you should not use the Mann-Whitney test if there are ties (1, 1, 1 in your case), as the test is not reliable.
However, most statisticians are OK to use this test if there are just a small number of ties.
Hope this helps,
Dr E.
Sound is very good!
Does anyone know when Mann Whitney last came up? I haven’t seen it in a psychology a level exam in years
Excellent...Well explained👏
I have a doubt here. When our table value is less than our calculated value, we reject the null hypothesis and vive versa. But in the video, table value is 5 and calculated value is less than 5 so why are we rejecting the null hypothesis
Hi Rupali,
We compare the U crit value (from the table) with the calculate U stat.
If the U stat is less than or equal to the U crit, the Null Hypothesis is rejected (as in the video).
If the U stat is greater than the U crit, we fail to reject Null Hypothesis.
Hope this helps,
Dr E.
Thanks.very easy to understand the way you taught us
why we have rejected the null hypothesis even we know that Ustat< Ucritical
you’re amazing ! thank you so so much
How do we take the medial value when the number occurs 3 times?
Exceedingly well explained
You are an absolute godsend xx
How would you write this hypothesis in scientific notation?
This was very useful. Thank you!
What pen do you use for writing..it is so neat and legible! Needless to say your explanations are on-point!
how is that...when stat value less than critical value and we reject null hypothesis....
in other test we fail to reject H0 in such cases....or this is the only case..??
Hi Ranjan,
As stated at 8:20 in the video, this is the way you the MWU test - it is different than otehr tests.
Dr E.
According to Wikipedia Ua+Ub should equal Na*Nb. in my data set i have 2 groups for a total of 126 responses. Na=67 Nb=59. Responses range between 0 and 10 in their value.
Group A has:
17 responses of 0 ranked at 11.5
3 responses of 1 ranked at 26
4 responses of 2 ranked at 33
5 respoonses of 3 ranked at 44
6 responses of 4 ranked at 55.5
20 responses of 5 ranked at 78.5
2 responses of 6 ranked at 99
3 responses of 7 ranked at 103
2 responses of 8 ranked at 110
0 responses of 9
and 5 responses of 10 ranked at 120.5
Group B has
5 responses of 0 ranked at 11.5
4 responses of 1 ranked at 26
3 responses of 2 ranked at 33
10 responses of 3 ranked at 44
2 responses of 4 ranked at 55.5
18 responses of 5 ranked at 78.5
1 responses of 6 ranked at 99
2 responses of 7 ranked at 103
7 responses of 8 ranked at 110
0 responses of 9
and 7 responses of 10 ranked at 120.5
So I calculated my Ua and Ub and tried to check whether I did it right, However, when i add Ua and Ub i get 3985, whilst Na*Nb amounts to 3953. So i must be doing something wrong. I have no clue what. Does anyone have any clue what is going wrong?
Happens almost fail-proof. You calculate something wrong, and you know it is but not why... as soon as you start panicking and flailing around until you give up and ask others for help you see your mistake...., Mine was that i calculated the U's wrong. I did Ua= Ra-((Na(Nb+1)/2) instead of Ua=Ra-((Na(Na+1)/2)..... Aaaanyway thanks a bunch for the vid, also helped me alot. Cheers from the Netherlands!
Thank you so much! Made it super easy for me.
Do you account for a 0 in the table
I am confuse which formula is use for u test
If calculated value is less than table value, the testing procedures says to accept Ho. But why you rejected ??
Sruthi K exactly!!!!!
@@harshithamajji9108 I looked it up on sphweb.bumc.bu.edu/otlt/mph-modules/bs/bs704_nonparametric/BS704_Nonparametric4.html and it says to reject