DEAR VIEWERS: I have a much newer video (ruclips.net/video/Ay5KbT_KrZo/видео.html) from February 2021 on this topic that was made with SPSS version 26. It's a higher quality video, and there you are able to download the dataset from the link underneath the video description. I hope you will consider visiting that video. Additionally, I have another video (Feb 2021) recently made on Kruskal-Wallis test here: ruclips.net/video/9EfmWS-YBDQ/видео.html . The Kruskal-Wallis test is effectively a generalization of the Mann-Whitney.
The U-test requires independent observations and assumes that the dependent variable is continuous. One consequence of this fact is that it is very unlikely for two individuals to have exactly the same score. This means that there should be few if any, tied scores in the data. When sample data do not have several tied scores, you should suspect that a basic assumption underlying the Mann-Whitney U-test has been violated. In this situation, you should be cautious about using the Mann-Whitney U.
thank u for the vide. By taking tied ranks and averaging those ranks, we get the assigned ranks. When doing this the result for the u of man withney must be different, my question is: why does the result remain the same (u and z and p) if we already have new assigned ranges?
Thank you for this very informative video, it has helped my in my Master's research thesis project. One quick question, my professor suggested I use the nonparametric tests ->independent samples to arrive at the mann-whitney u test instead of what you demo. why do you think that is?
Hi there. To tell you the truth I don't know why. There is some redundancy in some of the functions in SPSS (for instance, you can run one-way ANOVA via one-way anova under Compare Means and also run it using General linear model). However, the two approaches yield identical results (although the output and the options in your output might look different. As far as I am aware, there's no difference in the z value and asymptotic significance between the version I demo and the other route when performing Mann Whitney. Personally, I prefer this route more because I can also request mean ranks, allowing me to say more than the two groups are different (and instead talk about which group has the higher average rank on the dependent variable). Of course, you could still talk about the means of your two groups, even though the test is formally of the average of the ranks for your groups. You might also check out: statistics.laerd.com/spss-tutorials/mann-whitney-u-test-using-spss-statistics.php . I hope this helps.
@@mikecrowson2462 Thank you so much for the additional resources and your explanation, it definitely helped. I had to merge some of the original variables to fit my research design and when I took the route my professor suggested I actually kept getting an error code that would not display the average mean rank which I needed for my final results table (PESKY APA format). But you're route yielded all the data I needed and the yielded the same p-value results as the other way. I am glad to say I finished my results section and I can move onto my discussion. I will be sharing your page with all my fellow peers in confusion. Thanks again! You rock.
if p value is less then 0.05 ..there is a significant difference between two groups right ?i mean null hypothesis is not accepted...by the way can you please prove numurical formula of Mann whitney U test.. please..
Thanks for the video. Im having some issue. I have to perform the Mann-Whiney test using 8 value of protein concentration in plasma and 8 values of protein concentration in serum. Each pair of plasma and serum come from the same patient so there are 8 total patient samples. Im supposed to compare them and their difference. The problem is I dont have a grouping variable so I cant do the test. Do you have a tutorial of something similar that I can follow? thanks!
Hi Nikolai, I'm not sure what you mean when you refer to "comparing" protein and serum concentration. Is your question concerning how plasma and serum concentration within patient are related? I don't really have a good idea what you are seeking to test. Best wishes!
I am comparing 2 groups with continuous variables as outcomes (total # of units of medication). When entering the initial data I have average of 1000 units (+) per individual patient, but then my mean rank comes out to only 50's or 60's. How could it be? Am I doing something wrong?
Hi there, Saeed. The difference in sample size should not make a difference. The procedure essentially rank-orders values on the dependent variable across groups and then involves comparing the average of the ranks between groups. Not sure what to make out of the 'no exact sig value' issue. However, this is a way older video on this procedure and I would recommend checking out my newest video (2021) here (ruclips.net/video/Ay5KbT_KrZo/видео.html). There's also a link under the video description where you can download the data and follow along. Best wishes.
DEAR VIEWERS: I have a much newer video (ruclips.net/video/Ay5KbT_KrZo/видео.html) from February 2021 on this topic that was made with SPSS version 26. It's a higher quality video, and there you are able to download the dataset from the link underneath the video description. I hope you will consider visiting that video. Additionally, I have another video (Feb 2021) recently made on Kruskal-Wallis test here: ruclips.net/video/9EfmWS-YBDQ/видео.html . The Kruskal-Wallis test is effectively a generalization of the Mann-Whitney.
Thanks a world for your valuable information. You helped me a lot
It's my pleasure!
you saved my life!!!
Thank you. Very clear
The U-test requires independent observations and assumes that the dependent variable is continuous. One consequence of this fact is that it is very unlikely for two individuals to have exactly the same score. This means that there should be few if any, tied scores in the data. When sample data do not have several tied scores, you should suspect that a basic assumption underlying the Mann-Whitney U-test has been violated. In this situation, you should be cautious about using the Mann-Whitney U.
Thank you Sir
thank u for the vide. By taking tied ranks and averaging those ranks, we get the assigned ranks. When doing this the result for the u of man withney must be different, my question is: why does the result remain the same (u and z and p) if we already have new assigned ranges?
Thank you for this very informative video, it has helped my in my Master's research thesis project. One quick question, my professor suggested I use the nonparametric tests ->independent samples to arrive at the mann-whitney u test instead of what you demo. why do you think that is?
Hi there. To tell you the truth I don't know why. There is some redundancy in some of the functions in SPSS (for instance, you can run one-way ANOVA via one-way anova under Compare Means and also run it using General linear model). However, the two approaches yield identical results (although the output and the options in your output might look different. As far as I am aware, there's no difference in the z value and asymptotic significance between the version I demo and the other route when performing Mann Whitney. Personally, I prefer this route more because I can also request mean ranks, allowing me to say more than the two groups are different (and instead talk about which group has the higher average rank on the dependent variable). Of course, you could still talk about the means of your two groups, even though the test is formally of the average of the ranks for your groups. You might also check out: statistics.laerd.com/spss-tutorials/mann-whitney-u-test-using-spss-statistics.php . I hope this helps.
@@mikecrowson2462 Thank you so much for the additional resources and your explanation, it definitely helped. I had to merge some of the original variables to fit my research design and when I took the route my professor suggested I actually kept getting an error code that would not display the average mean rank which I needed for my final results table (PESKY APA format). But you're route yielded all the data I needed and the yielded the same p-value results as the other way. I am glad to say I finished my results section and I can move onto my discussion. I will be sharing your page with all my fellow peers in confusion. Thanks again! You rock.
Thanks a world for your valuable information.
Could you please explain me how to compare an ordinal variable between two groups in 3 times?
if p value is less then 0.05 ..there is a significant difference between two groups right ?i mean null hypothesis is not accepted...by the way can you please prove numurical formula of Mann whitney U test.. please..
Thanks for the video. Im having some issue. I have to perform the Mann-Whiney test using 8 value of protein concentration in plasma and 8 values of protein concentration in serum. Each pair of plasma and serum come from the same patient so there are 8 total patient samples. Im supposed to compare them and their difference. The problem is I dont have a grouping variable so I cant do the test. Do you have a tutorial of something similar that I can follow? thanks!
Hi Nikolai, I'm not sure what you mean when you refer to "comparing" protein and serum concentration. Is your question concerning how plasma and serum concentration within patient are related? I don't really have a good idea what you are seeking to test. Best wishes!
thnkx you..
You are very welcome. Cheers!
I am comparing 2 groups with continuous variables as outcomes (total # of units of medication). When entering the initial data I have average of 1000 units (+) per individual patient, but then my mean rank comes out to only 50's or 60's. How could it be? Am I doing something wrong?
I think you have dependent and independent variables the wrong way around for groupings - not 100% sure tho
Hi Alyssa, they are in the correct places in the menu box. Thanks for visiting!
Thank u for the helpful video.
My inquiry is if I have a control group (n=63) and a case group (n40.
Thanks
Hi there, Saeed. The difference in sample size should not make a difference. The procedure essentially rank-orders values on the dependent variable across groups and then involves comparing the average of the ranks between groups. Not sure what to make out of the 'no exact sig value' issue. However, this is a way older video on this procedure and I would recommend checking out my newest video (2021) here (ruclips.net/video/Ay5KbT_KrZo/видео.html). There's also a link under the video description where you can download the data and follow along. Best wishes.