Hello guys, could someone enlighten me and advise me in choosing a statistical test to evaluate responses from patients before and after a treatment, the variables being categorical (> 2 categories).
You can use paired sample t-test for your study. A paired sample t-test can be used to compare means of the same group before and after treatment.
Hello Victor,
The McNemar test is the best test for nominal variables with two dependent sample. When the categories are more than two, marginal homogeneity tests can be used. They are an extension of the McNemar test for dependent samples.
Good luck with your work!
Nominal
Eg. Vegetables
0= none
1= 1-2 days
2= 3-4 days
3= 5-6 days
4= 7 days
It´s in SPSS software.
In that case you can also use the Friedmann test which has already been incorporated in spss:
https://statistics.laerd.com/spss-tutorials/friedman-test-using-spss-statistics.php
Best,
Sven
The appropriate test for a more than 2 category is the Friedman test.
Data is ordinal in nature (> 2 categories) and before and after study so the Wilcoxon signed-rank test would be used.
https://statistics.laerd.com/spss-tutorials/wilcoxon-signed-rank-test-using-spss-statistics.php
Good morning Victor Guimarães. The variable you described is ordinal, not nominal.
- 0= none
- 1= 1-2 days
- 2= 3-4 days
- 3= 5-6 days
- 4= 7 days
Here is a web-page that offers advice on how to analyze paired ordinal variables:
- https://peterstatistics.com/CrashCourse/4-TwoVarPair/OrdOrd/OrdOrdPair0.html
I had never seen this site before, and cannot find any info on "Peter", the site creator. So I cannot vouch for it. But it might give you some ideas about how to proceed. I see that "Peter" also has a YouTube channel.
- https://www.youtube.com/c/stikpet/featured
HTH.
Hi dear Bruce Weaver, I've seen some articles like my paper and found: "Nutrition Behaviors in Polish Adults before and during COVID-19 Lockdown". I Would like to do such as that study, but they used McNeimar. However, it depends on the kind of variable nature.
Victor Guimarães, here is an article by Alan Agresti (1983) that you might find helpful:
- http://users.stat.ufl.edu/~aa/articles/agresti_1983.pdf
Cheers,
Bruce
Hi, Victor Guimarães . It would be helpful if you could answer some questions about your data to determine which analysis is appropriate in your case.
1) Do you have two times? Or more than two times?
2) Is your data such that you know what each individual responded for each time? That is, Person A responded X at Time 1 and Person A (the same person) responded Y at Time 2?
3) What is it you want to know?
a) In the example table you posted, you said they used McNemar test. This is okay. But it does treat the categories as nominal. It determines if there were differences the responses from one time to the other. But it _doesn't_ determine if the responses where higher or lower. McNemar doesn't know that "once a day" is greater than "once a week". It just knows that they are different.
b) Some of the suggestions in this thread recommend tests that would treat the categories as ordinal.
c) The difference here depends on what you want to know. For example, imagine you have ordered categories like in your vegetable example, and that we can label these categories 1 to 5 for convenience. Let's say that at Time 1 most people answered 3. (They eat vegetables moderately often). But at Time 2, for some reason, half of the people answer 1 (they rarely eat vegetables), and the other half answers 5 (they eat vegetables very frequently). McNemar will see this as very different from Time 1 to Time 2. But an analysis that treats the categories as ordinal may not see these as different. After all, the responses went from about 3 to about 3.
c1) On the other hand, if there was a subtle increase in the value of responses, an ordinal test might detect this, while a nominal test might not.
Hi, Sal Mangiafico I have two different moments (before and after the pandemic). In my data, I have values for each person in two moments.
I would like to know if there was a difference in the frequency of consumption before and after the pandemic. However, as you said, from the tests already mentioned I cannot determine which category is promoting this difference. There would be some supplementary test with which I could do this. According to what I have read, the marginal homogeneity test would be the most consistent for the type of analysis I want to do. What I didn't understand was why the article cited uses the MecNemar test and it uses more than 2 categories.
- Badges
- Science topic