I have a stats question...

I have one dependent variable and 2 independent variables...what kind of test do I run?

to be more specific; the one indepenent variable is glucose and the two independent variables are exercise and diet.


Thanks!

Edit: I was thinking 2-way ANOVA, but wanted to make sure

Depends one what you want to know and what the variables are (numeric, ordinal?).

You could do an ANOVA or a regression (both are based off the GLM).

The two-way ANOVA sounds correct. You'll need to think about what comparisons are appropriate though and whether they should be post hoc or a priori.

I think you meant to say glucose is the DV. Can you list the levels of the two IVs? Do you have any predictions about their effects, especially their interaction?

Well, I am writing a grant for a class, which is just an assignment.

But what I;m writing the grant on is my actual thesis topic.

I want to have two experimental interventions (diet or exercise) and it's acute effect on glucose levels.

It would be a crossover study, which leads me to believe that a dependent t test would work as well.

I was thinking of doing Tukey's post hoc test with it.

I don't think my results will vary from each other very much. But, I set my alpha at 0.05


And you are correct, i meant to say glucose was a dependent variable.

So you are assigning people to either diet or exercise? There is not a diet and exercise condition? Also, what do you mean by crossover study? Diet/exercise is a within subjects manipulation?

(Within meaning participants would be assigned to diet, for example. After a few weeks on diet, they would be assigned to exercise)

You aren't at A&M for grad school by chance are you?

No, I am in north texas for grad school

and yes

If it's a within subjects design (ie, participants do both conditions), then do a paired t test. If it's a between subjects design (ie, participants do only one of the conditions), then do an independent samples t test

Also interested to read what you mean by crossover study

Ya if you just have the two conditions a dependent or paired t will work. No control in your design?

Also, since its within subjects, are you counterbalancing the order?

crossover meaning half will do one experiment while the other half with do the other

a washout period will be had

then they switch....

You'll probably also want to examine order effects. I can imagine a scenario where they might occur with those two conditions.

Also, do you have a specific reason why you're doing this within instead of between subjects?

ehh, i'll worry about that when I actually set my thesis up (which will be next semester)

Well, just an fyi. If I imagine myself on your thesis committee that's probably a question I have. Members of your department may not care though.

Within subject could help answer the question oF does either intervention work better in some? Like are u genetically predispositioned to favor diet over exercise or vice versa

Interesting. Will you collect other data to examine this as a possible explanation? You could potentially do that between subjects as well if you conducted a mediational analysis (or moderational).

Really it depends on what your department expects in terms of research experience. Your advisor should help answer the question of how much you need to put into it.

I just presented this topic to him a couple of weeks ago. He OK'ed the topic, but there are some stuff we need to work through...

That clarifies it some... Are you going to measure your DV after each treatment or just at the end of the experiment?

after each treatment

It's been several years since I last did a study with a similar design so my thoughts on what to do may be inaccurate. I think the statistical test depends on what metric you use: their "performance" at each interval or if you use a difference score (performance at time A - performance at time B). I assume you'll do the former. In which case you probably want to do a 2x2 ANOVA

You're only measuring after each treatment? No baseline measure?