Stupid undergrad stats question -- help!

[amphigory]

I have what is probably a really stupid stats question but don't know where else to go, so I'm going to look stupid and hope someone can help me.

So I'm in this research methods class where we basically made up fake data and ran it in SPSS, and now I have to write a presentation/paper on it. The problem is that we didn't really get into post hoc analysis during the course, and I need it to interpret the data.

I have a mixed 2x4 design, where I'm comparing 4 groups on 2 conditions. I got significant main effects for group and condition, as well as an interaction of group*condition. At the last minute, my instructor ran a Tukey HSD which shows me the difference between the four groups. What I need to know is the post hoc analysis of the 2 conditions -- ie, on one condition, how are the four groups different? Does that make sense?

Can I run a one-way ANOVA between groups for each of my 2 conditions separately and then get the Tukey HSD for each condition?

I apologize if this is really dumb, I'm just trying to figure out how to do something I haven't been taught and was hoping I'm on the right track. Anyone? Thank you!

---

Members don't see this ad.

 

[JockNerd]

Well, how you do it would really depend on your research question, but what you're saying looks fine to me, if that's the comparison you want to make.

I'll just assume your data meets all the assumptions of running Tukey's HSD

 

[amphigory]

Thank you!

As for research question, basically my hypothesis was that the four groups would be the same in one condition, but different in the other condition. So knowing that overall the four groups are significantly different, and how they are different, is useful, but I also need to know if they are significantly different, and how, in each individual condition.

One other thing, while I'm feeling dumb. Is the mean difference significant at the .05 level if the p value is .051? I'm assuming not, but I want the answer to be "no" so I don't want that to color my judgment.

 

[Thrak]

One other thing, while I'm feeling dumb. Is the mean difference significant at the .05 level if the p value is .051? I'm assuming not, but I want the answer to be "no" so I don't want that to color my judgment.

Mathematically, no, it's not. Practically speaking, you could get away with calling it "marginally significant" since some journals tolerate that, but generally only if other results cross the .05 threshold.

Although if you hear my stats prof lecture on the topic, he's very opposed to .05 being some kind of "magical" number. But that's for a different topic

Thrak,
procrastinating at 3:00am when he should be finishing off a major project

 

[Ollie123]

Thrak is correct, but you should at the very least mention the value, don't just say "There was no significant difference (p > .05)" and move on. While it might be technically correct, its pretty sketchy to call something insignificant because it fits the hypotheses when p was .051. I would absolutely nail someone to the floor on a review if they did that, correct or not. One covariate you didn't include in analysis or didn't measure measure and that p is now well under .05. For an undergrad class though, I'm sure its fine.

What you described originally for the statistical procedure sounds fine to me unless I'm misunderstanding you.

 

[amphigory]

Thanks so much, guys. Paper and presentation finished!