RE: st: anova account for the same individuals

["HUANG, LILING" <HUANGL@mailbox.sc.edu>]

Dear David,

Thank you very much for your response to my inquiry. For this two-way analysis of variance, can I run pairwise comparisons to test the difference in rating scores between two ads, e.g., tukeyhsd? How do I specify this in the post-hoc test?
The scale of values for the rating scores is a 5-point Likert scale, ranging from 1 to 5. Indeed, this scale was created from 5 variables with values ranging from 1-5.  If the values are too few to run ANOVA, which method would be appropriate, xtmixed?
Thanks,

Liling
________________________________________
From: owner-statalist@hsphsun2.harvard.edu [owner-statalist@hsphsun2.harvard.edu] on behalf of David Hoaglin [dchoaglin@gmail.com]
Sent: Tuesday, November 19, 2013 6:04 PM
To: statalist@hsphsun2.harvard.edu
Subject: Re: st: anova account for the same individuals

Dear Llling,

Since each person rated all eight ads, an analysis of variance of your
scores should have two factors: ad and person.  It would be a two-way
analysis of variance.  If the 54 people constitute, in some reasonable
sense, a sample from some population, you would classify person as a
random factor. You probably are not considering the eight ads as a
sample from a population of ads (though that would be appropriate in
some instances), so you would classify ad as a fixed factor.

This ANOVA enables you to make comparisons among ads within person.
In that way, it generalizes the analysis underlying a paired t-test.

Stata may want you to treat the analysis as involving repeated measures:
anova score person ad, repeated(ad)
You can get further information from the documentation for -anova-.

You have not mentioned the scale of values for the rating scores.  If
the scale has only a few values, the ANOVA may not be entirely
appropriate.

I hope this helps.

David Hoaglin

On Tue, Nov 19, 2013 at 2:22 PM, HUANG, LILING <HUANGL@mailbox.sc.edu> wrote:
> Dear Statalisters,
>
> I have a group of people (n=54), rating eight different ads. So, I have 432 observations in a long format dataset. After I ran ANOVA on the rating scores among ads (F (7, 424) = 13.37, p<0.001), I used pairwise comparisons (tukey hsd) to test whether rating scores are significantly different between ads. My question now is how I can account for the ratings come from the same 54 people rather than from 432 people in this analysis since the ratings from the same individual (id) among ads are dependent. Please let me know how specify this in my syntax below. Or should I use other statistic method?
>
> anova score ad
> tukeyhsd ad
>
>
> Your prompt response to my inquiry will be greatly appreciated.
>
> Liling Huang

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/