Hi there,

I am interested in estimating the difference in (mean) between 30 students when they are on campus versus when they are outside (on holidays).  T

here is no clear dependent variable.  I am only interested in finding out if there are significant differences in mean frequency of drinking reported when students are on campus versus if they are on outside.

My dataset is set up as follows:

Student ID:     <====>  Drinking Frequecy-Internal <====> Drinking FrequencyExternal

Student01<====>24<====>35

I'm doing the same for a number of other variables, examining the difference in reported behavior while on campus, versus when they are off-campus. E.g, smoking frequency, number of nights spent out, etc.

All variables are on interval or ratio scale. At the end, i hope to produce a table that looks at each of the variables of interest and whether the difference in their means is significant. It will look like this:

Variable:       MeanOnCampus        MeanOffCampus      SigLevel for Difference

Smoking:
Drinking:
LateNight Parties:
etc....

Which tests will be appropriate to test difference in means - Anova, T-tests, or their non-parametric versions?

cheers......CY


---------------------------------
Do you Yahoo!?
 Yahoo! Small Business - Try our new resources site!

Paired t-tests will do what you want.  You're comparing means for TWO
groups of observations (on-campus vs. off-campus)-- hence t-test, not
ANOVA (which is for three or more groups).  Also, the two groups of
observations are not independent, as they would be if you were comparing
two sets  of students.  Instead you are comparing two different aspects
of one group of students -- so an  independent samples t-test (aka
UNpaired t-test) wouldn't be appropriate.  Check out PROC TTEST.

proc ttest;
        paired drinkon*drinkoff ;

Sarah Carroll, PhD
Research Coordinator
DHS - Immunization Branch, MS 7313
2151 Berkeley Way, Room 723E, Berkeley CA 94704
tel:  510.540.2484     fax:  510.883.6015
email:  scarroll@dhs.ca.gov


-----Original Message-----
From: SAS(r) Discussion [mailto:SAS-L@LISTSERV.UGA.EDU] On Behalf Of
Chao Yawo
Sent: Thursday, May 19, 2005 1:58 PM
To: SAS-L@LISTSERV.UGA.EDU
Subject: Testing Difference in Means for 30 cases


Hi there,

I am interested in estimating the difference in (mean) between 30
students when they are on campus versus when they are outside (on
holidays).  T

here is no clear dependent variable.  I am only interested in finding
out if there are significant differences in mean frequency of drinking
reported when students are on campus versus if they are on outside.

My dataset is set up as follows:

Student ID:     <====>  Drinking Frequecy-Internal <====> Drinking
FrequencyExternal

Student01<====>24<====>35

I'm doing the same for a number of other variables, examining the
difference in reported behavior while on campus, versus when they are
off-campus. E.g, smoking frequency, number of nights spent out, etc.

All variables are on interval or ratio scale. At the end, i hope to
produce a table that looks at each of the variables of interest and
whether the difference in their means is significant. It will look like
this:

Variable:       MeanOnCampus        MeanOffCampus      SigLevel for
Difference

Smoking:
Drinking:
LateNight Parties:
etc....

Which tests will be appropriate to test difference in means - Anova,
T-tests, or their non-parametric versions?

cheers......CY


---------------------------------
Do you Yahoo!?
 Yahoo! Small Business - Try our new resources site!

Actually, you could use a repeated measures ANOVA which is equivalent to
the paired t test when two observations are used. The ANOVA can be used
for more than two repeated measures as well.


Paul R. Swank, Ph.D.
Professor, Developmental Pediatrics
Medical School
UT Health Science Center at Houston

-----Original Message-----
From: SAS(r) Discussion [mailto:SAS-L@LISTSERV.UGA.EDU] On Behalf Of
Carroll, Sarah (DHS-DCDC-IMM)
Sent: Thursday, May 19, 2005 4:03 PM
To: SAS-L@LISTSERV.UGA.EDU
Subject: Re: Testing Difference in Means for 30 cases

Paired t-tests will do what you want.  You're comparing means for TWO
groups of observations (on-campus vs. off-campus)-- hence t-test, not
ANOVA (which is for three or more groups).  Also, the two groups of
observations are not independent, as they would be if you were comparing
two sets  of students.  Instead you are comparing two different aspects
of one group of students -- so an  independent samples t-test (aka
UNpaired t-test) wouldn't be appropriate.  Check out PROC TTEST.

proc ttest;
        paired drinkon*drinkoff ;

Sarah Carroll, PhD
Research Coordinator
DHS - Immunization Branch, MS 7313
2151 Berkeley Way, Room 723E, Berkeley CA 94704
tel:  510.540.2484     fax:  510.883.6015
email:  scarroll@dhs.ca.gov


-----Original Message-----
From: SAS(r) Discussion [mailto:SAS-L@LISTSERV.UGA.EDU] On Behalf Of
Chao Yawo
Sent: Thursday, May 19, 2005 1:58 PM
To: SAS-L@LISTSERV.UGA.EDU
Subject: Testing Difference in Means for 30 cases


Hi there,

I am interested in estimating the difference in (mean) between 30
students when they are on campus versus when they are outside (on
holidays).  T

here is no clear dependent variable.  I am only interested in finding
out if there are significant differences in mean frequency of drinking
reported when students are on campus versus if they are on outside.

My dataset is set up as follows:

Student ID:     <====>  Drinking Frequecy-Internal <====> Drinking
FrequencyExternal

Student01<====>24<====>35

I'm doing the same for a number of other variables, examining the
difference in reported behavior while on campus, versus when they are
off-campus. E.g, smoking frequency, number of nights spent out, etc.

All variables are on interval or ratio scale. At the end, i hope to
produce a table that looks at each of the variables of interest and
whether the difference in their means is significant. It will look like
this:

Variable:       MeanOnCampus        MeanOffCampus      SigLevel for
Difference

Smoking:
Drinking:
LateNight Parties:
etc....

Which tests will be appropriate to test difference in means - Anova,
T-tests, or their non-parametric versions?

cheers......CY


---------------------------------
Do you Yahoo!?
 Yahoo! Small Business - Try our new resources site!

thanks for all the responses.

I run both the t-test and the wilcoxon tests. Both found the same effects significant - there was such a close agreement. There were large differences between on and off campus cases though.

Out of 26 pairs of effects, about 20 of them were strongly significant with either test. But given the small number of cases, isn't this surprising ?

cheers, CY


Chao Yawo <yawo1964@yahoo.com> wrote:
Hi there,

I am interested in estimating the difference in (mean) between 30 students when they are on campus versus when they are outside (on holidays).  T

here is no clear dependent variable.  I am only interested in finding out if there are significant differences in mean frequency of drinking reported when students are on campus versus if they are on outside.

My dataset is set up as follows:

Student ID:     <====>  Drinking Frequecy-Internal <====> Drinking FrequencyExternal

Student01<====>24<====>35

I'm doing the same for a number of other variables, examining the difference in reported behavior while on campus, versus when they are off-campus. E.g, smoking frequency, number of nights spent out, etc.

All variables are on interval or ratio scale. At the end, i hope to produce a table that looks at each of the variables of interest and whether the difference in their means is significant. It will look like this:

Variable:       MeanOnCampus        MeanOffCampus      SigLevel for Difference

Smoking:
Drinking:
LateNight Parties:
etc....

Which tests will be appropriate to test difference in means - Anova, T-tests, or their non-parametric versions?

cheers......CY


---------------------------------
Do you Yahoo!?
Yahoo! Small Business - Try our new resources site!
__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around
http://mail.yahoo.com

Not really surprising, to me.  Students behave very differently on and
off campus.......well, yeah.  What might be more interesting
substantively is trying to find out why the differences in the different
behvaiors are different.....(what a phrase).....e.g. which behaviors
differ most and least between on- and off-campus.

This would require looking at effect sizes, not p-values - but you
shouldn't really be looking at p-values anyway......

It is nice that the Wilcoxon and t give very similar results.

Peter

Peter L. Flom, PhD
Assistant Director, Statistics and Data Analysis Core
Center for Drug Use and HIV Research
National Development and Research Institutes
71 W. 23rd St
www.peterflom.com
New York, NY 10010
(212) 845-4485 (voice)
(917) 438-0894 (fax)



>>> Chao Yawo <yawo1964@YAHOO.COM> 5/20/2005 2:09:36 PM >>>
thanks for all the responses.

I run both the t-test and the wilcoxon tests. Both found the same
effects significant - there was such a close agreement. There were large
differences between on and off campus cases though.

Out of 26 pairs of effects, about 20 of them were strongly significant
with either test. But given the small number of cases, isn't this
surprising ?

cheers, CY


Chao Yawo <yawo1964@yahoo.com> wrote:
Hi there,

I am interested in estimating the difference in (mean) between 30
students when they are on campus versus when they are outside (on
holidays).  T

here is no clear dependent variable.  I am only interested in finding
out if there are significant differences in mean frequency of drinking
reported when students are on campus versus if they are on outside.

My dataset is set up as follows:

Student ID:     <====>  Drinking Frequecy-Internal <====> Drinking
FrequencyExternal

Student01<====>24<====>35

I'm doing the same for a number of other variables, examining the
difference in reported behavior while on campus, versus when they are
off-campus. E.g, smoking frequency, number of nights spent out, etc.

All variables are on interval or ratio scale. At the end, i hope to
produce a table that looks at each of the variables of interest and
whether the difference in their means is significant. It will look like
this:

Variable:       MeanOnCampus        MeanOffCampus      SigLevel for
Difference

Smoking:
Drinking:
LateNight Parties:
etc....

Which tests will be appropriate to test difference in means - Anova,
T-tests, or their non-parametric versions?

cheers......CY


---------------------------------
Do you Yahoo!?
Yahoo! Small Business - Try our new resources site!
__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around
http://mail.yahoo.com