Message-ID: <e00f1f66-5d4c-db73-8615-aa030ecb79d3@mcmaster.ca>
Date: 2021-01-20T23:10:39Z
From: John Fox
Subject: Different results on running Wilcoxon Rank Sum test in R and SPSS
In-Reply-To: <22722_1611168756_10KIqYij027846_935783767.1030737.1611168336100@mail.yahoo.com>
Dear Bharat Rawlley,
On 2021-01-20 1:45 p.m., bharat rawlley via R-help wrote:
> Dear Professor John,
> Thank you very much for your reply!
> I agree with you that the non-parametric tests I mentioned in my previous email (Moods median test and Median test) do not make sense in this situation as they treat PFD_n and drug_code as different groups. As you correctly said, I want to use PFD_n as a vector of scores and drug_code to make two groups out of it. This is exactly what the Independent samples median test does in SPSS. I wish to perform the same test in R and am unable to do so.
> Simply put, I am asking how to perform the Independent samples median test in R just like it is performed in SPSS?
I'm afraid that I'm the wrong person to ask, since I haven't used SPSS
in perhaps 30 years and have no idea what it does to test for
differences in medians. A Google search for "independent samples median
test in R" turns up a number of hits.
>
> Secondly, for the question you are asking about the test statistic, I have not performed the Wilcoxon Rank sum test in SPSS for the PFD_n and drug_code data. I have said something to the contrary in my first email, I apologize for that.
For continuous data, the Wilcoxon test is, I believe, a reasonable
choice, but not when there are so many ties. If SPSS doesn't perform a
Wilcoxon test for a difference in medians, then there's of course no
reason to expect that the p-values would be the same.
Best,
John
> Thank you very much for your time!
> Yours sincerelyBharat Rawlley On Wednesday, 20 January, 2021, 04:47:21 am IST, John Fox <jfox at mcmaster.ca> wrote:
>
> Dear Bharat Rawlley,
>
> What you tried to do appears to be nonsense. That is, you're treating
> PFD_n and drug_code as if they were scores for two different groups.
>
> I assume that what you really want to do is to treat PFD_n as a vector
> of scores and drug_code as defining two groups. If that's correct, and
> with your data into Data, you can try the following:
>
> ------snip ------
>
> > wilcox.test(PFD_n ~ drug_code, data=Data, conf.int=TRUE)
>
> ??? Wilcoxon rank sum test with continuity correction
>
> data:? PFD_n by drug_code
> W = 197, p-value = 0.05563
> alternative hypothesis: true location shift is not equal to 0
> 95 percent confidence interval:
> ? -2.000014e+00? 5.037654e-05
> sample estimates:
> difference in location
> ? ? ? ? ? ? ? -1.000019
>
> Warning messages:
> 1: In wilcox.test.default(x = c(27, 26, 20, 24, 28, 28, 27, 27, 26,? :
> ? cannot compute exact p-value with ties
> 2: In wilcox.test.default(x = c(27, 26, 20, 24, 28, 28, 27, 27, 26,? :
> ? cannot compute exact confidence intervals with ties
>
> ------snip ------
>
> You can get an approximate confidence interval by specifying exact=FALSE:
>
> ------snip ------
>
> > wilcox.test(PFD_n ~ drug_code, data=Data, conf.int=TRUE, exact=FALSE)
>
> ??? Wilcoxon rank sum test with continuity correction
>
> data:? PFD_n by drug_code
> W = 197, p-value = 0.05563
> alternative hypothesis: true location shift is not equal to 0
> 95 percent confidence interval:
> ? -2.000014e+00? 5.037654e-05
> sample estimates:
> difference in location
> ? ? ? ? ? ? ? -1.000019
>
> ------snip ------
>
> As it turns out, your data are highly discrete and have a lot of ties
> (see in particular PFD_n = 28):
>
> ------snip ------
>
> > xtabs(~ PFD_n + drug_code, data=Data)
>
> ? ? ? drug_code
> PFD_n? 0? 1
> ? ? 0? 2? 0
> ? ? 16? 1? 1
> ? ? 18? 0? 1
> ? ? 19? 0? 1
> ? ? 20? 2? 0
> ? ? 22? 0? 1
> ? ? 24? 2? 0
> ? ? 25? 1? 2
> ? ? 26? 5? 2
> ? ? 27? 4? 2
> ? ? 28? 5 13
> ? ? 30? 1? 2
>
> ------snip ------
>
> I'm no expert in nonparametric inference, but I doubt whether the
> approximate p-value will be very accurate for data like these.
>
> I don't know why wilcox.test() (correctly used) and SPSS are giving you
> slightly different results -- assuming that you're actually doing the
> same thing in both cases. I couldn't help but notice that most of your
> data are missing. Are you getting the same value of the test statistic
> and different p-values, or is the test statistic different as well?
>
> I hope this helps,
> ? John
>
> John Fox, Professor Emeritus
> McMaster University
> Hamilton, Ontario, Canada
> web: https://socialsciences.mcmaster.ca/jfox/
>
> On 2021-01-19 5:46 a.m., bharat rawlley via R-help wrote:
>> ? Thank you for the reply and suggestion, Michael!
>> I used dput() and this is the output I can share with you. Simply explained, I have 3 columns namely, drug_code, freq4w_n and PFD_n. Each column has 132 values (including NA). The problem with the Wilcoxon Rank Sum test has been described in my first email.
>> Please do let me know if you need any further clarification from my side! Thanks a lot for your time!
>> structure(list(drug_code = c(0, 0, 1, 1, 1, 0, 0, 0, 1, 1, 0,?1, 0, 1, 0, 1, 1, 0, 0, 0, 1, 0, 1, 1, 0, 1, 0, 0, 0, 0, 0, 1,?0, 1, 0, 0, 1, 1, 0, 1, 1, 1, 1, 1, 0, 0, 1, 0, 1, 0, 0, 0, 1,?1, 0, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1, 0, 0, 1, 1, 0, 0, 0, 0, 1,?0, 0, 1, 1, 1, 0, 1, 1, 0, 1, 1, 1, 0, 0, 1, 1, 1, 0, 0, 1, 0,?1, 1, 0, 1, 1, 0, 1, 0, 0, 1, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0, 1,?1, 1, 1, 0, 0, 1, 0, 1, 0, 0, 1, 1, 1, 1, 0, 0), freq4w_n = c(1,?NA, NA, 0, NA, 4, NA, 10, NA, 0, 6, NA, NA, NA, NA, NA, 10, NA,?0, NA, NA, NA, NA, 0, NA, 0, NA, NA, NA, 0, NA, 0, NA, NA, NA,?NA, NA, NA, NA, NA, 0, 0, 12, 0, NA, 1, 2, 1, 2, 2, NA, 28, 0,?NA, 4, NA, 1, NA, NA, NA, NA, NA, 0, 3, 1, NA, NA, NA, NA, 4,?28, NA, NA, 0, 2, 12, 0, NA, NA, NA, 0, NA, 0, NA, NA, NA, NA,?NA, NA, NA, NA, NA, 3, NA, NA, NA, NA, NA, NA, 6, 1, NA, NA,?NA, 0, NA, NA, NA, 0, 0, NA, 0, NA, 2, 8, 3, NA, NA, NA, 0, NA,?NA, NA, 9, NA, NA, NA, NA, NA, NA, NA, NA), PFD_n = c(27, NA,?NA, 28, NA, 26, NA, 20, NA, 30, 24, NA, NA, NA, NA, NA, 18, NA,?28, NA, NA, NA, NA, 28, NA, 28, NA, NA, NA, 28, NA, 28, NA, NA,?NA, NA, NA, NA, NA, NA, 28, 28, 16, 28, NA, 27, 26, 27, 26, 26,?NA, 0, 30, NA, 24, NA, 27, NA, NA, NA, NA, NA, 28, 25, 27, NA,?NA, NA, NA, 26, 0, NA, NA, 28, 26, 16, 28, NA, NA, NA, 28, NA,?28, NA, NA, NA, NA, NA, NA, NA, NA, NA, 25, NA, NA, NA, NA, NA,?NA, 22, 27, NA, NA, NA, 28, NA, NA, NA, 28, 28, NA, 28, NA, 26,?20, 25, NA, NA, NA, 30, NA, NA, NA, 19, NA, NA, NA, NA, NA, NA,?NA, NA)), row.names = c(NA, -132L), class = c("tbl_df", "tbl",?"data.frame"))
>>
>> Yours sincerely?Bharat Rawlley? ? On Tuesday, 19 January, 2021, 03:53:27 pm IST, Michael Dewey <lists at dewey.myzen.co.uk> wrote:
>>
>> ? Unfortunately your data did not come through. Try using dput() and then
>> pasting that into the body of your e-mail message.
>>
>> On 18/01/2021 17:26, bharat rawlley via R-help wrote:
>>> Hello,
>>> On running the Wilcoxon Rank Sum test in R and SPSS, I am getting the following discrepancies which I am unable to explain.
>>> Q1 In the attached data set, I was trying to compare freq4w_n in those with drug_code 0 vs 1. SPSS gives a P value 0.031 vs R gives a P value 0.001779.
>>> The code I used in R is as follows -
>>> wilcox.test(freq4w_n, drug_code, conf.int = T)
>>>
>>>
>>> Q2 Similarly, in the same data set, when trying to compare PFD_n in those with drug_code 0 vs 1, SPSS gives a P value 0.038 vs R gives a P value?< 2.2e-16.
>>> The code I used in R is as follows -
>>> wilcox.test(PFD_n, drug_code, mu = 0, alternative = "two.sided", correct = TRUE, paired = FALSE, conf.int = TRUE)
>>>
>>>
>>> I have tried searching on Google and watching some Youtube tutorials, I cannot find an answer, Any help will be really appreciated, Thank you!
>>> ______________________________________________
>>> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
>>> https://stat.ethz.ch/mailman/listinfo/r-help
>>> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
>>> and provide commented, minimal, self-contained, reproducible code.
>>>
>>
>
> [[alternative HTML version deleted]]
>
> ______________________________________________
> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>
--
John Fox, Professor Emeritus
McMaster University
Hamilton, Ontario, Canada
web: https://socialsciences.mcmaster.ca/jfox/