anova on binary variables

[Deirdre Nissenson Kurzweil]

is it "legal" to use a one-way anova to check for significant
differences on a binary variable (e.g., gender)?

i'm being lazy and trying to do one big anova run for all my
dependent measures (separately, of course), rather than using t-tests
for some...

--
Deirdre Nissenson Kurzweil
Market Research Manager
Publications and Marketing
Cornell University
353 Pine Tree Rd.
Ithaca, NY 14850

Email: [hidden email]
Phone: [phone]

[Dr. Jeffrey D. Leitzel]

Hi Dierdre,
Any analysis that can be done with a t-test can be done w/ANOVA. The F
statistic from an ANOVA for a two group analysis will always equal the t
statistic squared. The significance results will be the same.
So, yes, it is "legal" and no statistical police will be issuing a warrant
for your arrest.
Best, Jeff
Jeffrey D. Leitzel, Ph.D.
Assistant Professor, Department of Psychology
Office: McCormick 2123
Bloomsburg University
400 East Second Street
Bloomsburg, PA  17815
Office Phone:[phone],fax:[phone]
Alt. Office (T & Th): 570 348-6100 ext:3216


-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Deirdre Nissenson Kurzweil
Sent: Tuesday, June 27, 2006 9:24 AM
To: [hidden email]
Subject: anova on binary variables

is it "legal" to use a one-way anova to check for significant
differences on a binary variable (e.g., gender)?

i'm being lazy and trying to do one big anova run for all my
dependent measures (separately, of course), rather than using t-tests
for some...

--
Deirdre Nissenson Kurzweil
Market Research Manager
Publications and Marketing
Cornell University
353 Pine Tree Rd.
Ithaca, NY 14850

Email: [hidden email]
Phone: [phone]

[statisticsdoc]

Stephen Brand
www.statisticsdoc.com

Deidre,

I assume that you are using Gender as an independent variable, not a dependent variable, that your dependent variables are interval scales, and that you are interested in knowing whether the means on these scales differ between males and females.  I would suggest conducting a t-test instead of the one-way ANOVA in the event that the variances on some of the dependent variables are not equal for males and females.    ANOVA assumes that the variances are equal across groups, while the t-test can be carried out without pooling unequal variances.   Using the t-test will give you more accurate results.

HTH,

Stephen Brand

---- Deirdre Nissenson Kurzweil <[hidden email]> wrote:

> is it "legal" to use a one-way anova to check for significant
> differences on a binary variable (e.g., gender)?
>
> i'm being lazy and trying to do one big anova run for all my
> dependent measures (separately, of course), rather than using t-tests
> for some...
>
> --
> Deirdre Nissenson Kurzweil
> Market Research Manager
> Publications and Marketing
> Cornell University
> 353 Pine Tree Rd.
> Ithaca, NY 14850
>
> Email: [hidden email]
> Phone: [phone]

--
For personalized and experienced consulting in statistics and research design, visit www.statisticsdoc.com