Questions/response
I received an email with the following questions:
"I know how to do every step except E, F, and G. In Regression, we would make a histogram and a QQ plot to test normal distribution. These are options we checked when we ran the analysis, but i do not see these options when we are running our ANOVAs. Additionally, how do you test the assumption of homogeneity of variance with ANOVAs? A number of us were in the lab today and we all had these questions."
My First suggestion for assumptions would be to re-check your notes (HINT: especially check Class One Handout). Is ANOVA robust to such assumptions? If you still have to check assumptions, remember there were multiple ways to get QQ plots and histograms, separate from the plots button. If you need further help with the rest of these, let us know...I would be glad to show you in further detail...but i have a feeling you'll have an "AHA" moment!
Any other suggestions Mari?
"I know how to do every step except E, F, and G. In Regression, we would make a histogram and a QQ plot to test normal distribution. These are options we checked when we ran the analysis, but i do not see these options when we are running our ANOVAs. Additionally, how do you test the assumption of homogeneity of variance with ANOVAs? A number of us were in the lab today and we all had these questions."
My First suggestion for assumptions would be to re-check your notes (HINT: especially check Class One Handout). Is ANOVA robust to such assumptions? If you still have to check assumptions, remember there were multiple ways to get QQ plots and histograms, separate from the plots button. If you need further help with the rest of these, let us know...I would be glad to show you in further detail...but i have a feeling you'll have an "AHA" moment!
Any other suggestions Mari?
posted by KGL @ 6:19 PM
3 Comments:
At 10:37 PM,
Mari said…
Besides post all such questions to the blog and don't email the TAs with them? : )
Actually, I think Katie has answered these questions pretty well.
At 9:23 AM,
Nikki Frederick said…
So, it sounds like if Anova is robust, we do not need to do additional things to check the assumptions? This doesn't feel correct to me, but that's what Katie's response implies to me???
At 8:12 PM,
Mari said…
ANOVA will check the assumption of homogeneity of variance for you (if you ask it to do so).
Katie has noted what you should do to check the assumption of normality.
The assumption of independence of observations is a design issue...not one that is formally tested. So, for instance, you will need to decide if you have set up your experiment in such a way that you have independence of observations (that is, do you have nonindependent observations such as you would have with husband-wife data)?
With this assignment, you have been granted control over the IV and DV, so its completely reasonable that you also have control over independence of observations as well. :)
Having said all this...although you should check the assumptions, it is also true that ANOVA is robust to minor violations of these assumptions. Should you find that the assumptions were violated, you would need to temper your conclusions...but conclusions belong in a Discussion section, not in a Results section.
Post a Comment
<< Home