Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: st: My ANOVA and regression results don't agree
From
"Pepper, Jessica" <[email protected]>
To
"[email protected]" <[email protected]>
Subject
RE: st: My ANOVA and regression results don't agree
Date
Wed, 8 Jan 2014 21:00:16 +0000
Hi David,
The outcome variable has a range of 1 to 5 and is very skewed (mean 2.04, SD 1.18, skewness 1.02). I do have balance of assignment of conditions (i.e. the type of ad) within sex.
The interaction term (adcontent#sex) is not significant in either the ANOVA or the regression. And when I drop the interaction from the model, the regression and ANOVA match again (i.e. both models show significant main effects for both variables). Given that situation, dropping the interaction from the model may be the way to go.
Jess
-----Original Message-----
From: [email protected] [mailto:[email protected]] On Behalf Of David Hoaglin
Sent: Wednesday, January 08, 2014 1:45 PM
To: [email protected]
Subject: Re: st: My ANOVA and regression results don't agree
Jess,
Your description of the boxplots is encouraging, but you have not given much information about your outcome variable (e.g., range of possible values, such as 0 to 100). You didn't comment on skewness, so it may be reasonable to use ANOVA or regression.
If you used random assignment separately for the women and the men, you should at least have balance (or near-balance) within sex. That structure, however, is not the usual 2-way ANOVA, which has the same number of observations in each cell. Thus, in the ANOVA you would want to look at the mean square for adcontent after you have accounted for variation associated with sex. The corresponding regression "automatically" adjusts for the contribution of sex.
I don't recall from earlier messages whether the interaction of adcontent and sex is significant. The analysis is simpler if you can omit the interaction from your model. Then you can compare the effects of the levels of adcontent after adjusting for the contribution of sex (e.g., look at differences among the coefficients for adcontent). If the interaction is significant, you are limited to comparing the effects of adcontent within each sex separately.
David Hoaglin
On Wed, Jan 8, 2014 at 11:46 AM, Pepper, Jessica <[email protected]> wrote:
> The boxplots show similar distributions (median, size of quartile, and max value) for 4 of the 6 combinations. The other 2 are similar to one another.
> Also, my data are not balanced; although there was random assignment to ad type, more women took the survey than men. In other aspects of my analysis (survey results), I am weighting my results to reflect this, but I was not planning to do this in the experiment.
> Do these 2 aspects (distribution of DV and lack of balance) change the ANOVA/regression situation?
*
* 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/