There is at least one mistake in this video - ranks should be assigned based on the ABSOLUTE VALUE of the difference, NOT the signed difference. So 12 should be rank 1, -14 and 14 should both be ranked 2.5, 16 is ranked 4, etc.
I'll give it another look. I don't perform this test very often, and the way I did it here may be based on a flawed textbook. Thanks for letting me know.
You use either the sum of the positive ranks, or the sum of the negative ranks, whichever is smaller. Maybe you're looking at an example from the book where the positive ranks were smaller?
@statslectures No it really states "The sum of positive ranks" and in the example the positive ranks are higher, in another example the positive ranks are lower, in both examples they take the sum of positive ranks.
They test a Median agains a Benchmark, does that matter?
first of all thank you very much for the video, very useful. I have a question, I tried to do your test with SPSS but it is giving me Z= -1.682 (Analyse>Nonparametric test>legacy Dialogs>2 related samples). can you please tell me what I am missing?
Very good and clear video. would have been nice if it was shown in a commonly used progem such as Excel as i got into a bit of a fix when trying to put that formula in excel. How do you get specific P-values from the Z-value, is that in an earlier video?
Thanks. In the future I hope to make videos for statistical programs such as Excel, SPSS, SAS, and R. I find the specific P-values by looking them up in a Z-table, which you see around the 1:00 mark.
Hi again, i posted a question in your page. I have scores from two different systems, e.g. system I, graded 3, 4, 3. and system II graded 2, 4, 2. I want to compare if two systems correlates well. Wilcoxon Signed-Ranks Test seems a good fit after I understood, since its taking the ranking not the values.. What do you think? Thanks in advance.
@statslectures I think this doesn't work for my problem. The reason is, all positive differences are not differentiated.. Sum R+ remains same.. Any ideas? Sorry for bothering..
The test in this video is actually to see if the two groups differ. If you want to see if they're correlated, you want to use a different kind of test. If you're correlating regular quantitative data you usually want to use a Pearson's r correlation, while if you're dealing with ordinal (ranked) data you usually want to use a Spearman's r correlation. I'd have to know more about the data to choose the correct test, but check out those two videos first.
Hi again, i have another question, hope you don't mind. I have scores from two different systems, e.g. system I, graded 3, 4, 3. and system II graded 2, 4, 2. I want to compare if two systems correlates well. More specially, the second score was higher in both cases and first and third was same.. So they correlates well, even though the differences are not equal. If there is another system with lower differences that should be ranked higher. Which test do you think will be appropriate..
Comment removed
colman77 3 days ago
There is at least one mistake in this video - ranks should be assigned based on the ABSOLUTE VALUE of the difference, NOT the signed difference. So 12 should be rank 1, -14 and 14 should both be ranked 2.5, 16 is ranked 4, etc.
colman77 3 days ago
Comment removed
colman77 3 days ago
@colman77
It really depends on who you ask and what your professor or teacher tells you to do, unfortunately.
statslectures 3 days ago
@statslectures
Take a look at the Wikipedia page; especially the External Link to cornell.edu. It includes a great explanation for why this test works.
with absolute value - this test approximates the probability the a dataset is centered around your test mean value.
without - ... tests for strange cases depending on whether len(data) is even or odd.
There are ongoing debates about a number of topics in statistics but I don't think this is one of them. Professors occasionally make mistakes.
colman77 3 days ago
@colman77
I'll give it another look. I don't perform this test very often, and the way I did it here may be based on a flawed textbook. Thanks for letting me know.
statslectures 3 days ago
@statslectures yeah it's a strange but interesting little corner of statistics. Thanks for taking another look, and thanks for the great videos!
colman77 2 days ago
My book says to calculate the Wilcoxon test statistic, use the SUM OF POSITIVE RANKS.
I'm lost.
AuronNU 3 months ago
@AuronNU
You use either the sum of the positive ranks, or the sum of the negative ranks, whichever is smaller. Maybe you're looking at an example from the book where the positive ranks were smaller?
statslectures 3 months ago
@statslectures No it really states "The sum of positive ranks" and in the example the positive ranks are higher, in another example the positive ranks are lower, in both examples they take the sum of positive ranks.
They test a Median agains a Benchmark, does that matter?
AuronNU 3 months ago
how should I perform Wilcoxon's test for one sample..eg: one sample data compared to a hypothesized median value..? Thanks in advance
ajaylamba31 7 months ago
first of all thank you very much for the video, very useful. I have a question, I tried to do your test with SPSS but it is giving me Z= -1.682 (Analyse>Nonparametric test>legacy Dialogs>2 related samples). can you please tell me what I am missing?
R0cknr 7 months ago
Very good and clear video. would have been nice if it was shown in a commonly used progem such as Excel as i got into a bit of a fix when trying to put that formula in excel. How do you get specific P-values from the Z-value, is that in an earlier video?
goldtoleadalchemist 11 months ago
@goldtoleadalchemist
Thanks. In the future I hope to make videos for statistical programs such as Excel, SPSS, SAS, and R. I find the specific P-values by looking them up in a Z-table, which you see around the 1:00 mark.
statslectures 11 months ago
Hi again, i posted a question in your page. I have scores from two different systems, e.g. system I, graded 3, 4, 3. and system II graded 2, 4, 2. I want to compare if two systems correlates well. Wilcoxon Signed-Ranks Test seems a good fit after I understood, since its taking the ranking not the values.. What do you think? Thanks in advance.
naushadzaman 1 year ago
This has been flagged as spam show
@statslectures I think this doesn't work for my problem. The reason is, all positive differences are not differentiated.. Sum R+ remains same.. Any ideas? Sorry for bothering..
naushadzaman 1 year ago
@naushadzaman
The test in this video is actually to see if the two groups differ. If you want to see if they're correlated, you want to use a different kind of test. If you're correlating regular quantitative data you usually want to use a Pearson's r correlation, while if you're dealing with ordinal (ranked) data you usually want to use a Spearman's r correlation. I'd have to know more about the data to choose the correct test, but check out those two videos first.
statslectures 1 year ago
Hi again, i have another question, hope you don't mind. I have scores from two different systems, e.g. system I, graded 3, 4, 3. and system II graded 2, 4, 2. I want to compare if two systems correlates well. More specially, the second score was higher in both cases and first and third was same.. So they correlates well, even though the differences are not equal. If there is another system with lower differences that should be ranked higher. Which test do you think will be appropriate..
naushadzaman 1 year ago
very good work.. can you please refer to the video where you explained how to get the value from Z table?
naushadzaman 1 year ago
@naushadzaman
There should be a pretty good example of me using the z-Table in this video:
"One Sample z-Test"
You should be able to find it from my YouTube channel (YouTube isn't letting me link it directly).
statslectures 1 year ago
@statslectures Hi. Heartiest thanks. Found it.. It was very useful.
naushadzaman 1 year ago