Pepijn Hesselink Both of your variables (gender, health status) are nominal, so a t-test is not appropriate. A binomial logistic regression will work. Or Pearson chi-square test, which is just a special case of the former.

Thank you for your response. A fisher's exact test should not be necessary in this case?

Also I need to do the same for other variables (age, body mass index, mean time since last meal). Since these are not nominal variables, an independent sample t test would be the best option, I assume?

Given those sample sizes, a Chi-square test of association (also called a test of independence) ought to be fine. For your example, it will test the null hypothesis that the proportion of males (or females if you prefer) does not depend on group.

You've not said what stats package(s) you have access to, but any of them will have a routine for carrying out this test. And if you don't have a stats package, there are nice online calculators, like this one:

• http://vassarstats.net/newcs.html

HTH.

Pepijn Hesselink Since you have a larger than 2 x 2 contingency table (3 x 2), Fisher's approach does not apply without modification. It will be simpler to use Pearson Chi Square. Both of your variables are categorical, so a t-test will not work. T-test requires that your dependent variable is continuous.

Pepijn Hesselink , I think your question has mislead commentators. In your comment, you say you have a collection of variables. Is your goal to explore is this collection of variables differs among those three groups? I am also assuming that this collection of variables are associated with each other. If this is true, it would be better to start a new question.

Daniel Wright I'm sorry if my comment was confusing. I have had barely any experience with statistical analysis and therefore have a hard time wrapping my head around this and making my questions comprehensible.

I need to know if there any statistical differences per variable. I need to do this seperately for 3 comparisons. So group 1 (controls) vs. group 2 (disease A), group 1 vs. group 3 (disease B) and group 2 vs. group 3.

These are the variables I want to test: gender ratio, age, body mass index, coffee consumption (yes or no), alcohol consumption (yes or no), tobacco consumption (yes or no).

I assume that a Chi-square test is sufficient for testing the gender ratio and coffee, alcohol and tobacco consumption variables.

I believe that an independent sample t test would be fitting for the age and body mass index variables.

I hope this clears my question up somewhat. Please let me know if I would need to ask a new question with a better explanation.

Pepijn Hesselink , thank you for the clarification. One issue is to examine the associations between these, so for example tobacco and alcohol consumption tend to be related in most populations. Things like age and BMI will likely have odd distributions, so even if one of these was your only variable t-test might not be the best choice. You will also want to deal with the three constrasts you list (1 v 2, 1 v 3, and 2 v 3) are not independent.

I guess first question is how were people sampled?

Daniel Wright Thank you for the information. I will elaborate on the research a bit more and hope that clears up everything with regards to the sampling method.

In short, the study concerned an electronic device that analyses exhaled breath particles. A pilot study demonstrated that the device was able to succesfully differentiate healthy controls and multiple sclerosis (MS) patients based on this breath analysis.

The follow-up study wanted to test if they could increase the sensitivity of the device and whether the device can differentiate between multiple types of MS (relapse-remitting MS and progressive MS). The breath analysis was used to create 3 seperate models (one model that compared the breath analysis of relapse-remitting MS to controls, one model that compared progressive MS to controls and the final one compared relapse-remitting MS and progressive MS.

Therefore I have my 3 groups. A control group, a relapse-remitting MS patient group and a progressive MS patient group who all participated in the breath analysis and filled in a questionaire in order to get the data such as age and gender, etc.

Now I want to determine if there any statistical differences that could influence the models that were created.

After writing this I can imagine that this might make things harder to fully comprehend. I hope it at least gives some further insight.

Does the breath device presumably gives some breakdown of the breath. Does this then give probabilities of being in these three groups? Do you have some other variable that provides some gold standard for which of these groups the person is in? Let's suppose you do (if not, you have other issues), then you presumably are asking about the relationship between this gold-standard classification and the breath analyses, and whether this relationship is moderated by the other variables (like gender, age, alcohol). Or is this something else?

Patients who were included in the MS patient groups had a confirmed diagnosis of either relapse-remitting or progressive ms. Controls consisted simply of people not suspected to have ms.

I indeed want to know if the other variables could have an influence on the reliability of the models and therefore I need to know if there are statistical differences between the comparison groups.

So this is different from your original question. I recommend deleting this one and starting a new question. Otherwise it will be confusing for others who search for help for their questions.

Based on the information you have provided, the test that seems most appropriate is the Chi-Square Test of Independence, also known as the Chi-Square Test of Association, which is a nonparametric test of significance for categorical variables. You might find the libguide by Kent State University an insightful read.

Kent State University. (2021, July 8). SPSS tutorials: Chi-square test of independence. https://libguides.library.kent.edu/spss/chisquare

Good luck,