Kruskal-Wallis-Test (Simply explained)
HTML-код
- Опубликовано: 6 июл 2024
- The Kruskal-Wallis test (H-test) is a hypothesis test for multiple independent samples, which is used when the requirements for a one factor analysis of variance are not met.
Since the Kruskal-Wallis test is a nonparametric test (also called a distribution-free procedure), the data used do not have to be normally distributed, in contrast to analysis of variance. The only requirement is that the data be ordinal scale.
In the Kruskal-Wallis test, ordinal variables are sufficient because in nonparametric procedures, it is not the differences of the values that are used, but the rank places (which value is larger, which is smaller). Therefore, the Kruskal-Wallis test is also often called rank variance analysis according to Kruskal and Wallis.
Calculate the example directly with DATAtab for free, just click on the link:
datatab.net/statistics-calcul...
More Information about the Kruskal-Wallis-Test:
datatab.net/tutorial/kruskal-...
And the Kruskal-Wallis-Test calculator:
datatab.net/statistics-calcul...
![Friedman Test [Simply Explained]](http://i.ytimg.com/vi/2moNzzkkZwU/mqdefault.jpg)
![Friedman Test [Simply Explained]](/img/tr.png)
![Chi-Square Test [Simply explained]](/img/1.gif)
If you like, please find our e-Book here: datatab.net/statistics-book 😎
Best statistical tutorials for those who want to study it. Thank you very much for your help.
This was so clear, thank you! I've been struggling to understand this concept for a while now.
You explain the information so well. Thank you!
Wonderful presentation, thank you for simplifying the explanation!
Glad it was helpful!
Calculate the example directly with DATAtab for free, just click on the link:
datatab.net/statistics-calculator/hypothesis-test/kruskal-wallis-test-calculator?example=kruskal_wallis_test_reaction_time
Thank you so much. I'm not fluent in English, but I can understand your explanation really well. It's very helpful. I have a question, how to do Kruskal-Wallis Test on sample data of plant leaf growth that were observed at 1-week intervals. Is the Kruskal-Wallis Test conducted for each observation (such as the first observation, second observation, etc.) or is it conducted only once with the data made one from all observations from the first to the end?
This was such an easy to follow example. Thank you so very much!!
Glad it was helpful!
Nice presentation. The tutorial is a great help.
This is the best explanation I have found for this test yet, thank you so much! The only thing I would add in an updated version is the solving of a full question from start to finish (as a university student, questions usually indicate that normality is not guaranteed, and leave the rest to you. It would be nice to see the process of completing this test in questions with a specific provided scenario.)
Hi many thanks for your feedback! That is a great point! I will have a look at it. Regards Hannah
Fantastic video, great work!
Many thanks for the nice Feedback! Regards Hannah
I am curious about why chi-squared can be used at the end. When reading about chi-squared, it says it is only suitable for a large sample size (about N>50). But Kruskal-Wallis-Test can be used even with a much smaller sample size. My guess is chi-squared is just an approximation of the actual distribution (H distribution?), but I am not very sure.
Thank you so much! This tutorial helps me greatly!
Glad it helped!
Hello! Thank you for your video. You said that the expected value equation is when there is no difference in the group? Do you mean the same number of samples in each group? What if there is different number of samples in each group? Thank you
hello! if i have 13 variables and i want to know how to check which one is significant and which one is not?
Thank you so much for the video! I'm currently running this test with ordinal categories (salary range) and number of right answers on a test (discrete variable), is this appropriate?
Many thanks! Yes this sound correct! Regards Hannah
Thank you for clear explanation
This is an excellent, clear and concise explanation
Glad it was helpful and many thanks for your feedback! Regards Hannah
Thank you so much! It was very helpful!
You're very welcome! Many thanks for your nice Feedback! Regards Hannah
Awesome ! Very helpful. Thank you!!
Glad it was helpful!
Which test can i use if my groups are not mutually exclusive?
Thanks for your wonderful explanation
Many thanks! Regards Hannah
I was struggling to understand this test
Thank you so much for this simple and easy video !
Glad it helped!
Very nice comprehensive explanation
Glad you liked it and many thanks for the nice feedback! Regards Hannah
beautifully explained! thanks
Glad it was helpful!
Thanks for the informative video!
Can the Kruskal-Wallis test be used instead of a 2-way ANOVA when assumptions have been violated? If not, which test would you recommend?
Best James
The Kruskal-Wallis test is more closely aligned with a one-way ANOVA, as it's used to compare three or more independent groups on a single, continuous dependent variable. If you're dealing with a situation that would normally require a two-way ANOVA (involving two independent variables), the Kruskal-Wallis test wouldn't be the direct equivalent because it doesn't accommodate the analysis of interaction effects between two independent variables. Unfortunately there is no non-parametric counterpart, at least not that I know of! Regards Hannah
Thank you very much for online lesson
Thanks for your Feedback! Regards Hannah
Which test should I use when I have an independent variable with 2 levels and dependent variable with 3 levels?
Maybe Chi2-Test? datatab.net/tutorial/chi-square-test
thanks for the great video. how can I do kruskal -wallis for the 60 column and 40 raw?
With datatab : ) datatab.net/statistics-calculator/hypothesis-test/kruskal-wallis-test-calculator
@@datatab Thanks
Is it free? I could not get result, I think it is not free software. right?
perfect job!
Many thanks! Regards Hannah
Thanks for your time and teaching Madam
It's my pleasure
may i ask, if the data are not normally distributed, is it the null hypothesis should be rejected?
Hello, I think I do not quite understand the question unfortunately! The normal distribution has nothing to do with the null hypothesis. Regards Hannah
THANK YOU! I should use this test for an Ordinal (1-5 from a Likert scale) Depend Variable, with Independet nominal variables with more than 2 levels, right?
Yes exactly! It may also be that the Likert scale can be assumed to be metric and normally distributed, then you could even take an ANOVA!
@@datatab thank you! But what if the Levene() test comes out not and there's no homogeinity?
Can we also use paired ANOVA ?? Because I am working on Spss and after inserting data I checked normality test and it says that it's parametric
A repeated Mesures ANOVA is the parametric copunterpart to the Kruskal Wallis Test. So if all assumptions are meet you can use the rm ANOVA. Regards Hannah
Hi! at 6:59, for rank variance, is the 12 in the denominator due to the number of cases you have, or will 12 always be used? Great video btw
Hi, 12 will always be there! Regards, Hannah
Can run this test for measure relationship in continous and catergarical variable
In this case you will use the Friedman Test: The Friedman test analyzes whether there are statistically significant differences between three or more dependent samples. Here is my Video for the Friedman Test ruclips.net/video/2moNzzkkZwU/видео.html
Very well explained. Thankyou Mam
Most welcome 😊
Helpful, thanks! So it's like the Mann-Whitney test, but for more than 2 groups.
Glad it was helpful! Yes exactly! Regards Hannah
Hi, I noticed that your equation is different from the the one I found online( with 3(N+1)), why is that so?
Tell me if u found the answer
What is the non-parametric test for 2 way and higher factorial anova?
Oh sorry, I can't tell you off the top of my head either! Maybe you use a generalized regression model with ranks and there is no separate name for the method, but I do not know exactly!
I don’t know about whether there even exists a non parametric alternative for 3-way ANOVA or higher, but for 2-way balanced factorial ANOVA, it’s called the Schierer Ray Hare test.
@@woodchuk1 Great! Many thanks!!!! Regards, Hannah
Gutes Video :) Es gibt einen Versprecher bei 2:15 min (Ranksum 42 gesagt statt 54). Danke für Euren Content!
Oh danke für dein Feedback!!!! LG Hannah & Mathias
Can this test be used to determine normal distribution?
hmm I don't think so!
Thank you so much ma'am ❤
Thanks!
Thank you so much. I have understood😊
You're welcome 😊
Crystal clear
This saved my life
Greate!
can we use the same method when we have tie ranks? If so, how do we rank the ties?
Thank you.
You're welcome!
Madam, Kruskal Wallis is non parametric test compared to ANoVa. Once there is group difference later we go for post hoc to observe which group is different from others. For kruskal Wallis, in data tab, is there feature available for post hoc ? Thanks for your time 🤲🤲🌹🙏
Does this H formula still hold even if there are ties in ranks?
When you have tied ranks the calculation is a little different! Regards Hannah
@@datatab Different how?🙂
How were we going to approach it if there were ties?
@@rakthestudentactuary5077 Unfortunately I haven't prepared anything for that yet, maybe it will come soon! I can't answer it here in the comments, just google it : )
Count how many of each type of tie there are (2-way tie, 3-way tie, etc.) and assign each tie a weight. An “n-way” tie for the purposes of the correction has weight n^3-n. Add up all the weights and divide by x^3 - x, where x is the total number of data points. Subtract this fraction from 1 to get the correction factor, then divide the calculated test statistic by the correction factor.
For example, say we have 25 data points in all that have between them 3 2-way ties and 2 4-way ties in the ranks. A 2-way tie has a weight of 2^3 - 2, or 6, and we have 3 of them, so that adds up to 3*6, or 18. The 4-way ties each have a weight of 4^3 - 4, or 60, and there are 2, so that makes 2*60, or 120. The sum is 18 + 120, or 138. So, divide 138 by (25^3 - 25) to get (138/15600) and subtract this from 1 to get (15462/15600). Divide your calculated test statistic by (15462/15600) to get the corrected value.
Thanks very helpful
Thanks for your nice feedback! : )
Thank you
You're welcome
Job well done
Many thanks! Regards Hannah
Refered book?
you are too good
Thanks : )
What are the indicators of if null hypothesis is not rejected
Hi, thanks for your question, what do you mean by indicators?
The question been answered around last part thanks datatab for replying!
@@user-bu1gf7wn2o Greate! Regards Hannah
Won't the null-hypothesis be rejected in this case? I think you said it will be retained or in other words, not rejected. The critical value is shown to be 5.901 while the H-value comes out to be 0.5; doesn't that mean that the chance of our observation as a result is less than 5% had the null-hypothesis been true, and hence it should be rejected?
Thank you in advance.
No. That H-stat corresponds to the Chi^2 stat. We are reading from a table of critical values and finding that our H stat would have to exceed 5.901 to yield a p-value below our alpha threshold of .05
superb, please give the video for "BOX PLOTS" in english.
Thanks for the Feedback, the english version will be there next week!
What is the P-value?
Please watch my video about the p-Value: ruclips.net/video/CL9MsExcKfU/видео.html
@@datatab Can you explain the expected value with more detail? I do not get it what is the explanation in this video.
It is not chi squar
It is called kai square
Many thanks for your Feedabck! In the latest videos we have paid attention to it! Regards Hannah
🖤
: )
Useful video.
I will always be a little uncomfortable hearing a German (edit: accent) discussing methodologies of dividing people into groups.
???
@@datatab Just a dark joke...
"chi" is pronounced wrong the whole time I have to quit the video...
Literally thought the same thing haha