Comparing Means

[Andrew Piskorowski]

All-

 

I have a dataset that contains scale responses. What is the proper
analysis for comparing the means of a subset of a group to be compared
to the whole group? And how would I access that in SPSS 16?

 

The two real problems are this is a non-independent analysis as the
group means contain the subset cases and what the proper comparison
would be for this type of data.

 

Any help would be greatly appreciated. Thanks.

 

-Andrew Piskorowski

 

====================To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Swank, Paul R]

This has been discussed before on one of my many list serves but it
continues to rear its head. The problem is that you are mixing dependent
and independent observations. For truly dependent observations you can
use a paired t test. For independent groups, you use an independent t
test. What I would suggest is to subdivide your sample into the subgroup
of interest and everybody else. Then use the independent samples t test.

Paul R. Swank, Ph.D.
Professor and Director of Research
Children's Learning Institute
University of Texas Health Science Center - Houston


-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Andrew Piskorowski
Sent: Thursday, July 24, 2008 1:47 PM
To: [hidden email]
Subject: Comparing Means

All-



I have a dataset that contains scale responses. What is the proper
analysis for comparing the means of a subset of a group to be compared
to the whole group? And how would I access that in SPSS 16?



The two real problems are this is a non-independent analysis as the
group means contain the subset cases and what the proper comparison
would be for this type of data.



Any help would be greatly appreciated. Thanks.



-Andrew Piskorowski



==========To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Andrew Piskorowski]

Thanks for your response Paul. Though, I'm unable to subdivide the
groups.

Essentially it's like comparing the accident rate for a group of people
who drive red cards to the accident rate of everyone who drives a car.
Those red car drivers would be included in everyone who drives a car and
it's just not possible to take them out of the population.

Or if you were comparing the mean heart rate of a group of high school
athletes  all from one school who came in for a physical to the mean
heart rates of all the high school athletes over the last 5 years.
Chances are some from the recent group are also in the 5 year sample.

Does that make sense?

Thanks.

-Andrew Piskorowski


-----Original Message-----
From: Swank, Paul R [mailto:[hidden email]]
Sent: Thursday, July 24, 2008 3:03 PM
To: Andrew Piskorowski; [hidden email]
Subject: RE: Comparing Means

This has been discussed before on one of my many list serves but it
continues to rear its head. The problem is that you are mixing dependent
and independent observations. For truly dependent observations you can
use a paired t test. For independent groups, you use an independent t
test. What I would suggest is to subdivide your sample into the subgroup
of interest and everybody else. Then use the independent samples t test.

Paul R. Swank, Ph.D.
Professor and Director of Research
Children's Learning Institute
University of Texas Health Science Center - Houston


-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Andrew Piskorowski
Sent: Thursday, July 24, 2008 1:47 PM
To: [hidden email]
Subject: Comparing Means

All-



I have a dataset that contains scale responses. What is the proper
analysis for comparing the means of a subset of a group to be compared
to the whole group? And how would I access that in SPSS 16?



The two real problems are this is a non-independent analysis as the
group means contain the subset cases and what the proper comparison
would be for this type of data.



Any help would be greatly appreciated. Thanks.



-Andrew Piskorowski



==========To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[zstatman]

Sure does though I don't agree that the subgroup should also be included in
the overall. After all, you want to compare the sub to the "not-sub" as I
see it and how Paul explains.

WMB
Statistical Services

============
mailto: [hidden email]
http:\\home.earthlink.net\~statmanz
============

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Andrew Piskorowski
Sent: Thursday, July 24, 2008 4:23 PM
To: [hidden email]
Subject: Re: Comparing Means

Thanks for your response Paul. Though, I'm unable to subdivide the groups.

Essentially it's like comparing the accident rate for a group of people who
drive red cards to the accident rate of everyone who drives a car.
Those red car drivers would be included in everyone who drives a car and
it's just not possible to take them out of the population.

Or if you were comparing the mean heart rate of a group of high school
athletes  all from one school who came in for a physical to the mean heart
rates of all the high school athletes over the last 5 years.
Chances are some from the recent group are also in the 5 year sample.

Does that make sense?

Thanks.

-Andrew Piskorowski


-----Original Message-----
From: Swank, Paul R [mailto:[hidden email]]
Sent: Thursday, July 24, 2008 3:03 PM
To: Andrew Piskorowski; [hidden email]
Subject: RE: Comparing Means

This has been discussed before on one of my many list serves but it
continues to rear its head. The problem is that you are mixing dependent and
independent observations. For truly dependent observations you can use a
paired t test. For independent groups, you use an independent t test. What I
would suggest is to subdivide your sample into the subgroup of interest and
everybody else. Then use the independent samples t test.

Paul R. Swank, Ph.D.
Professor and Director of Research
Children's Learning Institute
University of Texas Health Science Center - Houston


-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Andrew Piskorowski
Sent: Thursday, July 24, 2008 1:47 PM
To: [hidden email]
Subject: Comparing Means

All-



I have a dataset that contains scale responses. What is the proper analysis
for comparing the means of a subset of a group to be compared to the whole
group? And how would I access that in SPSS 16?



The two real problems are this is a non-independent analysis as the group
means contain the subset cases and what the proper comparison would be for
this type of data.



Any help would be greatly appreciated. Thanks.



-Andrew Piskorowski



==========To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command SIGNOFF SPSSX-L For a list of
commands to manage subscriptions, send the command INFO REFCARD

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command SIGNOFF SPSSX-L For a list of
commands to manage subscriptions, send the command INFO REFCARD

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Bob Walker-2]

[Maguin, Eugene]

Maybe the question that needs an answer is what constitutes that larger
group. It seems to me that depending on the answer to this question, the
recommendation would vary. If the larger group is a large random sample,
e.g., a 10% sample of California residents with a drivers license, and the
smaller, comparison sample is senior students with a drivers license in the
Sedro Woolley high school, then constructing a one sample test would make
sense. It's not perfect but it would seem to me that even if you did take
the smaller sample out of the larger sample the difference between leaving
them and taking them out would be less than decimal dust. The other endpoint
is where the smaller sample is a significant fraction of the larger sample.
Then I'd agree with other responders.

However, isn't it also true that if the total sample N, mean and SD is known
along with the N, mean, and SD of a subsample of the the total, then the N,
mean and SD of the total-subsample can be computed?

Gene Maguin

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Juanito Talili]

Hi all,
 
Using exploratory factor analysis(principal component analysis using varimax rotation) the 15 items were reduced to three factors.  Out of curiosity, using the same data I subjected the three factors to confirmatory factor analysis(CFA).  The CFA showed that one of the three factors was statistically nonsignificant (factor coefficient=.02; p>.05).  Why it was so?
Please help.
 
J Talili




====================To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Swank, Paul R]

It's not exactly clear what you mean by the one factor was nonsignificant. Did you mean that one indicator was not significant for one factor? You must remember that CFA is a much more restricted model than EFA. IN EFA, you can have non zero (albeit small) loadings on all factors whereas in CFA, you generally force most of the coefficients to zero. Thus, many times the CFA model will fail even though it was the EFA solution.

Paul R. Swank, Ph.D.
Professor and Director of Research
Children's Learning Institute
University of Texas Health Science Center - Houston


-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Juanito Talili
Sent: Thursday, July 24, 2008 6:49 PM
To: [hidden email]
Subject: EFA then CFA

Hi all,

Using exploratory factor analysis(principal component analysis using varimax rotation) the 15 items were reduced to three factors.  Out of curiosity, using the same data I subjected the three factors to confirmatory factor analysis(CFA).  The CFA showed that one of the three factors was statistically nonsignificant (factor coefficient=.02; p>.05).  Why it was so?
Please help.

J Talili




=======
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD