Mann-Whitney U test on SPSS with write up (APA style)

[@DavidRobinsonPhD]

Mann-Whitney U tests and many other analyses are covered in my book, SPSS Made Easy: www.amazon.co.uk/SPSS-Made-Easy-Statistical-Researchers/dp/B0DJGR4Z5K

[@Jennifer-hs5ps]

Ahhhh, apparently this is the channel to go to for interpretation and reporting analyses! Thanks again!

[@DavidRobinsonPhD]

Hi Jennifer, glad to help!

[@hazeldemaeijer2685]

Hi - just wanted to say massive thank you for this video. I'm doing a write up for my MSc and doing a mix of correlations and Mann Whitney U, which I had never done before,...I was feeling major stress about how I was going to tackle something new on a program I don't really have much love for at the moment, as well as interpret the output, and then how to write it all up in APA formatting, but this short video gave me EVERYTHINGI needed to smash out my results section for that part in one go and really reassured me that it wasn't too complicated! So thank you :)

[@DavidRobinsonPhD]

Hi Hazel, thanks your message, glad the video helped!

[@elisabethks9773]

I never write in the comment section of youtube! But my guy you helped me so muchhhh you can't even imagine what i went through to understand how this whole think works and you just made it so simple thank you so much!!!!!!!!!!!1

[@DavidRobinsonPhD]

Thanks Elisabeth, glad it helped!

[@dwara6358]

Your explanation is very clear. Thank you so much. Now I can do my analysis part without having any doubt 😊

[@DavidRobinsonPhD]

Thanks, glad it helped!

[@zsanettkukucska8213]

thank you so much - most decent explanation video on youtube in this topic lol

[@DavidRobinsonPhD]

Thanks, glad it helped!

[@pharohmonk]

Excellent video. Thanks for the explanation.

[@DavidRobinsonPhD]

Thanks Alan!

[@sandram.3305]

Thank you! Regards from Bosnia and Herzegovina.

[@DavidRobinsonPhD]

Thanks Sandra, glad the video helped!

[@marianoezequielzalazarcarr8927]

Hi david! (I am the guy from Argentina who greeted you earlier in the video of the logistic regresion). My question is about the following: I have carried out the analysis U of mann whitney, the p value is .002, but when it comes to checking the medians, both groups have the same result, as I interpret that?

[@DavidRobinsonPhD]

Hi Mariano, thanks for your question. There's an explanation of this here: stats.oarc.ucla.edu/other/mult-pkg/faq/general/faq-why-is-the-mann-whitney-significant-when-the-medians-are-equal/

[@jamiehenry7081]

Thanks so much 🙏 this was soooo helpful!

[@DavidRobinsonPhD]

Thanks Jamie, glad it helped!

[@hayleyspurin8286]

Really excellent explanation, thank you.

[@DavidRobinsonPhD]

Thanks Hayley, glad it helped!

[@riinatuokko4922]

Thank you very much for your advice. I made it! ☺️👍☀️

[@sharikzamir4373]

very informative and helpful :) 💯

[@wenboguo2238]

Thank you!!

[@amrotarrisi2658]

thank you very much , very much informative

[@DavidRobinsonPhD]

Thanks Amro, glad it was helpful!

[@sidrakanwal2097]

Hi ,I amhave a question if the grouo have different distribution in tow group three we have to report median rank,can you please let us know how to interpret median rank

[@DavidRobinsonPhD]

Hi Sidra, thanks for your question. Higher ranks reflect higher scores.

[@sidrakanwal2097]

@@DavidRobinsonPhD I have one interpretation that means rank of change of bmi in control is high and intervention groups is low. Does this mean change n bmi in control was high??

[@DavidRobinsonPhD]

@@sidrakanwal2097 If the mean rank for change scores is higher in the control condition than in the intervention condition, this indicates that, overall, change scores are higher in the control condition. However, this could be misleading if you have negative values in your dataset.

[@Fenix1445]

Hi David. What do you mean by this? "(If there's a minus symbol before your Z value, you can ignore it.)"

[@DavidRobinsonPhD]

Hi Luis, thanks for your question. For example, if the z value in SPSS is -3.22, you can ignore the minus symbol so that you divide positive 3.22 by the square root of N to get a positive r value. This is recommended in, e.g., the SPSS Survival Manual, I suppose because it’s more conventional to report positive effect size statistics.

[@joseluisalves5829]

@@DavidRobinsonPhD Thank you for your reply David. So for that calculation with the square root to find the r value I will discard the minus. What about when reporting apa style? can the minus also be discarded?

[@DavidRobinsonPhD]

@@joseluisalves5829 Couldn't see anything in the APA publication manual about this specifically. However, I would opt not to include the minus symbol as long as the direction of the effect has been established (e.g., you've stated which group has the higher median).

[@andrrrrew]

thank you!!!

[@HauntedBlack]

Thank you for this video! Question: Can you use the formula for r when you have unequal sample size (example: 40 and 18)? And why do you need report medians?

[@sarmisthaboruah3972]

How can i calcuate the IQR for each category/option as i want to report the IQR along with median

[@faridabnb5227]

Hello David, I have a question if you don't mind, i have pre test scores and post scores which means after my treatment group took the treatment,, the control group didn't take the treatment,but had both tests and the treatment group had both tests and the treatment in between so here , should i do the mann whitney test with post test scores only for both groups ?to see any significant difference between the groups Second ,when you counted the r value the z value was in your table -3.215 but you took only 3.215 , it was

[@DavidRobinsonPhD]

Hi Farida, thanks for your questions. Yes, this test could be used to complete between-group comparisons post-treatment. It could also be used to check that there weren't between-group differences pre-treatment. Regarding the r value, the video description might help to answer your question.

[@faridabnb5227]

@@DavidRobinsonPhD thanks a lot for your answer

[@faridabnb5227]

@@DavidRobinsonPhD hello David , please can you help ,i have a question,, my post test data is not normally distrbuted and i have an outlier that i did not remove, so i used the Mann Whitney u test to compare ,however my pre test data are normally distrbuted, so i wanted to use the same test to see any difference betwen my treatment and control groups before the treatment period . Is it ok for the type of test with pre test data? The sample is 46 students ,26 treatment and 21 control,the variances are equal for the pre test.

[@DavidRobinsonPhD]

@@faridabnb5227 Hi Farida, Thanks for your question. If you used a Mann-Whitney test for the post-intervention data, to ensure consistency I would probably also use it for the pre-intervention data even if they meet the assumptions for a t-test. You can find a discussion about this subject here: www.researchgate.net/post/Can_I_use_the_T-Test_and_the_Mann-Whitney_in_the_same_analysis Best wishes, David.

[@faridabnb5227]

@@DavidRobinsonPhD thank you so much for your answer ,i will mention in my study ,as you said, that i used Mann Whitney u test for pre test data for the sake of consistency .

[@pradeeparanaweera873]

Nice video, thank you. if the data set contained several zeros, How do conduct this test?

[@Kayla-dr5hv]

Hi David, is there any reason that my variable labels don't automatically change from a number in the data view tab to the label I created in the variable view tab? I have tried the SINTAX method too - no luck. Any help would be greatly appreciated, thank you!

[@DavidRobinsonPhD]

Hi Kayla, Thanks for your question. If you go to View then click Value Labels, your numbers should be replaced by labels. Best, David.

[@tsheringyenji3941]

Hi I have a question... Can the data set be non-significant for Mann-whitney U test. After running Man whitney test my p value is greater than .05. Can i still interpret this result?

[@DavidRobinsonPhD]

Hi, thanks for your question. Yes, that just means that there's not a significant difference between the groups.

[@tsheringyenji3941]

@@DavidRobinsonPhD Thank you sir for your quick response. I have one more question..... which table do we use to report our analysis. I want add a table in my paper, so do we use the Ranks table or Test Statistics table generated by SPSS? Thank you.

[@tbbek17]

the effect size of .72 - did you use Cohens D? how do you derive that it was a large effect when Cohen's D large effect is 0.8 (sorry very confused).

[@DavidRobinsonPhD]

Hi Tracy, Thanks for your question. In this case the effect size value is an approximate r value rather than a d value, and the criteria are different. For example, while a d value normally needs to be above .8 to be described as "large", an r value only needs to be above .5. Hope that helps! David.

[@Tiaaa2010x]

@@DavidRobinsonPhD Hi David, what if my effect size is .28 as this is lower than .5, do I still report this?

[@DavidRobinsonPhD]

@@Tiaaa2010x I would still report it, though would describe it as "small" rather than "large" (see Cohen's criteria in the description).

[@Tiaaa2010x]

@@DavidRobinsonPhD perfect thank you :)

[@lucinena9478]

If I want to make pies which number (means or media) need I use?

[@IslandRides]

Hi Luci, Thanks for your question. If you're referring to pie charts, these aren't usually used to represent means or medians, but rather how many (or what percentage of) people are in a certain group or responded a certain way. Best wishes, David.

[@lucinena9478]

@@IslandRides I measure motor ability in norms and I thought of representing with bar graphs. Thank you very much for the answer. I will wait anxiously for this too

[@DavidRobinsonPhD]

@@lucinena9478 If you measured motor ability in two groups, you could represent the medians (or means) for each group in a bar graph.

[@lucinena9478]

​@@DavidRobinsonPhD ​ :) thank you so much...

[@fobioma5545]

This was extremely helpful! You have a new subscriber here :) Any idea how to present the results in a table? I have several analysis to run and it would take too much space to write it all out.

[@DavidRobinsonPhD]

Hi Ihuoma, Thanks for your question. A few people have asked about tables recently, so will be sure to include them in future videos. Here's a good example of a Mann-Whitney table: www.researchgate.net/figure/Mann-Whitney-U-test-for-significance-of-social-vulnerability-for-highly-engaged-and-or_tbl2_303096277 Let me know if you have any questions. All the best, David.

[@fobioma5545]

@@DavidRobinsonPhD Thank you very much!

[@aisya1240]

Hi if the r number is negative do we consider it has a different between two groups too?

[@DavidRobinsonPhD]

Hi Aisya, Thanks for your question. Yes, whether it's positive or negative doesn't really matter in the context of this test. I would just report the r value without the minus symbol. To determine whether there's a difference between the conditions, focus on the p value (aka sig value). Regardless of whether the r value is positive or negative, the difference between conditions can be considered significant if the p value is below .05. Hope that helps! David.

[@aisya1240]

@@DavidRobinsonPhD Very clear and right to the point!Thank you so much Dr for the answer. Your videos really help a lot. Thank you for being such an amazing researcher and helping so many people out. 💖

[@DavidRobinsonPhD]

@@aisya1240 Thank you!

[@elinagailite6954]

Sometimes medians are equal even when Mann-Whitney shows significant difference. This makes me think that one cannot make any conclusions by using only medians. Any comments how data should be interpreted? Which figures shall be used in interpretation?

[@DavidRobinsonPhD]

Yes, normally the p value is the most interesting value in the output. As you allude to, the medians might look quite similar or quite different, but usually what we're interested in is whether the p value suggests a significant difference; medians could seem quite distinct but the difference could still be non-significant as sample size affects the likelihood of a difference being significant.

[@dimitrachorianopoulou7958]

the volume is always too low in your videos

[@DavidRobinsonPhD]

Thanks for the feedback Dimitra.