Simeon Stoynov , if new variable gender and activity level is created then we will have four groups namely; active men, inactive men, active women and inactive women. Afterwards, the four groups can be compared through Kruskal Wallis test.

Simeon Stoynov, you don't need to worry. By grouping variable, SPSS is asking you to enter, in your case, the variable that you have coded for your participants' sex.

Once you enter that variable into the field for grouping variable, you'll get a new option in which you will, presumably, enter 1 for males and 2 for females - or vice versa.

Thanks for the quick answer Robert Trevethan . I specifically need to compare VO2 Max scores between physically active women vs physically inactive men. The Mann Whitney U test did allow me to define the sex grouping variable (which I did as 0 and 1) but from what I'm seeing, the test results are just separating VO2 Max scores and physical activity entirely for men and women and isn't showing me what I need to see.

EDIT: I should clarify that in my data, there are some men and women who are physically active, and men and women who are physically inactive.

Simeon Stoynov, something seems to be going wrong. Ah, I think I've worked it out. You want to conduct your comparison ONLY between the men and women who are physically active. If so, you need to select them at the start so that only those people are included in your MWT.

So, go to Data, Select Cases, and work with the fields that open up there. If you're not sure, the following site might help:

https://www.youtube.com/watch?v=NUSam9mRiV0

Let me know how you go!

Thanks for the quick reply again Robert Trevethan

I need to specifically compare physically active women vs physically INactive men. I should probably have been a bit clearer with my terminology.

Simeon Stoynov, my apologies: I didn't read your question carefully enough.

You can still go into Data, Select Cases - but then you need to do some careful nifty things with the IF statements. If you're not sure how to do it, I know there are some useful YouTube sites showing how.

Again, sorry for my oversight.

Robert Trevethan I've been tinkering with it and I was just about to come back to say that your previous answer with the video helped me a lot. The filtering function allows the filtering of multiple variable combinations, in my case, letting the data display -only- physically active women -and- physically inactive men. Thank you.

Robert Trevethan Sir, I am learning many things after going through your answers. Thank you for sharing your knowledge with us.

Simeon Stoynov, you're welcome. Again, my apologies for not being sufficiently attentive at the start. Feel free to get back with any queries.

Incidentally, I wonder why you're using an MWT rather than a t-test for VO2 Max - but that's a whole other issue. In my experience, anyway, MWTs and t-tests usually yield similar results, so maybe it's not a big deal.

All the best with your research.

You provide a vague hypothesis:

"to compare the VO2 Max scores between physically active women and physically inactive men ."

Can you say what you are either comparing or testing? Without guessing no one can provide an answer.

Shirshendu Roy, I'm glad my posts were helpful.

I confess to feeling a little ashamed because I didn't read Simeon's question clearly enough initially - but sometimes good things can come out of mistakes.

Daniel Wright, you have made me realise that Simeon's question might be more complex than initially appeared. Thanks.

Simeon, do you want to just test (for some reason that you've not disclosed) whether the two groups of people differ in terms of VO2 Max, or do you want to determine the comparative extent to which sex and physical inactivity, as separate variables, are related to VO2 Max?

Those two kinds of question are approached differently in terms of statistical analysis.

I'm pretty sure the former could be addressed by either an MWU test or by a t-test. The latter could be addressed with multiple regression.

Once more, I'm sorry if I wasn't on the ball in my earlier advice. I was thrown off by your saying you wanted to use a nonparametric test - which I now suspect was a red herring.

Perhaps you could explain why you thought you needed to use a nonparametric test. That might help to clarify things. Maybe you could add how you are measuring physical inactivity. Presumably VO2 Max could be regarded as an equal interval variable - yes?

Simeon Stoynov As far as I undestand you have max VO2 scores from a group of people measured only once. You want to find out if physical activity and gender affect max VO2 scores(do you?).

If this is the case, then you can perform a two way ANOVA with two factors physical activity and gender. And have a look at the interaction which shows if activity changes have different effects on genders. By this way, you can also make comparisons between genders and activity levels.

I believe non-normality issue you are concerning will not refrain you from doing ANOVA.

Very enlightening answers from all, thank you.

Simeon Stoynov, it's courteous to thank those of us who have tried to help you, but I'm afraid we might have made things more complex for you than you'd anticipated.

Conducting statistical analyses is often that way, particularly when a question is not framed clearly or with sufficient information.

When you have time / inclination, I think it would be good if you answered the questions I asked in my most recent post above here. That includes how you are measuring VO2 Max. Have you permitted it to be an equal-interval (or at least continuous) variable, or is it, say, dichotomous?

If you answer that, it would be helpful to determine whether Mehmet's suggestion that you use two-way ANOVA is appropriate.

Robert Trevethan

I just ran the data through a Levene's test for equal variances - and from what I understand, the test showed that the variances are indeed equal (null hypothesis not rejected?). So this allows the use of a two-way ANOVA even though the data is not normally distributed as I understand it?

VO2 Max was measured once for all men and men with the same protocol. When this data was collected in the lab, it was not collected with a rigid hypothesis in mind. We were told (my course) to do with it as we like. I chose to compare physically active women to physically inactive men as some research suggests that men have a higher VO2 Max than women who are more physically active than them. In that case, I suspect that my data will show the same thing. I wanted also to see how sex and physical activity, as two independent variables, affect VO2 Max scores, the interactions of the variables as Mehmet Sinan Iyisoy brought up.

Simeon Stoynov, AOK with regard to your VO2 Max data. It seems you've measured it as a continuous (and I suspect equal-interval) variable.

How about physical (in)activity, however? How did you measure it? Was it on some kind of scale, or something such as just yes/no?

Simeon Stoynov , if new variable gender and activity level is created then we will have four groups namely; active men, inactive men, active women and inactive women. Afterwards, the four groups can be compared through Kruskal Wallis test.

Robert Trevethan It was a simple yes/no in regards to physical activity. Either you are or are not. I classified that as an ordinal variable, since 'yes' is more than 'no' in this scenario. I've just run the data through a two-way ANOVA (dependent variable being VO2 Max scores, fixed factors being sex -and- physical activity) and am in the process of trying to interpret it.

Both the test of homogeneity and test of equality of error variances came out good - null hypothesis not rejected.

Muhammad Mumtaz Khan Thank you for the answer, yes you are right about the four groups. Presently based on people's previous replies I think that I satisfy the enough of the requirements for a two-way ANOVA so I'll probably be going for that to compare two groups - active women and inactive men. I'll check out Kruskal Wallis if I decide I need to compare all four groups.

Simeon Stoynov, now that I can see what your variables are, I'm pretty sure that a two-way ANOVA is indeed the path to take.

There are some good YouTube presentations that might help you interpret your results, including interaction effects. Here's one:

https://www.youtube.com/watch?v=6Hmauags3VA

Robert Trevethan

I've just run a two-way ANOVA like this:

https://i.gyazo.com/610ee3952969bc49b3cf96ed1c6fa71d.png

However I'm having problems. The results of the tests of between-subjects effects looks like this, with lots of missing values especially in Sig. https://i.gyazo.com/d8d4eb5cb58982993fda085e4c87c6ba.png

The 38 total participants in this two-way ANOVA have no missing values in their variables. I'm not sure what is happening, but I ran the test again and again and it shows the same thing.

Did you include all observations into calculation? You may have a filter in effect.

Also make sure you have included factors into the model using model button.

Mehmet Sinan Iyisoy Thank you for the reply. There is indeed a filter in place, a filter that only leaves physically active women and physically inactive men to be tested. Since these are the two groups that I want to directly compare.

I just checked and I had not included factors into the model. I am not sure how this is done. https://i.gyazo.com/166b9d6fa4fed134ff99ed36c4c660a4.png

Should the 'Type' be interaction as it's currently set? What about the rest of the settings?

Simeon Stoynov, sorry for not being available immediately. It looks as if you still have the filters engaged that I suggested you use when I thought you had only two groups that you wanted to compare. So, you need to go back to Data / Select Cases - and choose all cases!

I might not have the exact choices indicated there, but I'm pretty sure that something like that will work.

Robert Trevethan No worries you have been invaluable in this discussion. I've removed the filter and while the test results look more reasonable now, this no longer specifically compares physically inactive men to physically active women as I originally wanted. This is what I see now:

https://i.gyazo.com/b4e4b6c34caaf135af7981bdfcdb482c.png

Is there any workaround or any course of action I could take after this?

It seems OK now. Now only aerobic training seems effective on max VO2 taking also sex into account. Sex and interaction effects are not significant. So for both sexes, you may have significant differences of max VO2 in different training groups. I am not sure if you can compare physically inactive men with active women with this setup.