Before and after experiment using factor analysis

[kimberley2314]

Hello,

I'm doing an experiment where I'm measuring consumers attitude before and after seeing an advertisement. For the questions before and after the advertisement I'm using the same scale that measures two dimensions. However, when I'm doing an factor analysis on those 2 questions I can extract 2 factors in each question, but they don't consist of the same items...
My question is if I can compare the corresponding factors...? (the corresponding factors do measure the same dimensions but they don't consist of the same items after conducting the factor analysis)

Kind regards,
Kimberley

[Swank, Paul R]

This means your scale is not invariant over time. It could be that the intervention plays a role in causing the change but it does mean that any change over time is confounded with a change in the construct definition.

Dr. Paul R. Swank,
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 kimberley2314
Sent: Thursday, April 28, 2011 8:23 AM
To: [hidden email]
Subject: Before and after experiment using factor analysis

Hello,

I'm doing an experiment where I'm measuring consumers attitude before and
after seeing an advertisement. For the questions before and after the
advertisement I'm using the same scale that measures two dimensions.
However, when I'm doing an factor analysis on those 2 questions I can
extract 2 factors in each question, but they don't consist of the same
items...
My question is if I can compare the corresponding factors...? (the
corresponding factors do measure the same dimensions but they don't consist
of the same items after conducting the factor analysis)

Kind regards,
Kimberley

--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Before-and-after-experiment-using-factor-analysis-tp4346460p4346460.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
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

[Gosse, Michelle]

Hi Kimberley,

I see Paul gave you good advice in his reply.

What is the purpose/hypothesis of your study? You've done an experiment; I would expect that at least one of your hypotheses is that the advertisement affected consumer attitudes - so I'm not seeing why you would expect the same factor (principal component?) loadings to occur before and after advertisement exposure.

Cheers
Michelle

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of kimberley2314
Sent: Friday, 29 April 2011 1:23 a.m.
To: [hidden email]
Subject: Before and after experiment using factor analysis

Hello,

I'm doing an experiment where I'm measuring consumers attitude before and
after seeing an advertisement. For the questions before and after the
advertisement I'm using the same scale that measures two dimensions.
However, when I'm doing an factor analysis on those 2 questions I can
extract 2 factors in each question, but they don't consist of the same
items...
My question is if I can compare the corresponding factors...? (the
corresponding factors do measure the same dimensions but they don't consist
of the same items after conducting the factor analysis)

Kind regards,
Kimberley

--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Before-and-after-experiment-using-factor-analysis-tp4346460p4346460.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
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

**********************************************************************
This email and any files transmitted with it are confidential and
intended solely for the use of the individual or entity to whom they
are addressed. If you have received this email in error please notify
the system manager.

This footnote also confirms that this email message has been swept by MIMEsweeper for the presence of computer viruses.
**********************************************************************

=====================
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

[kimberley2314]

First of all, thanks for the answers!

The scale that I'm is using is a hedonic/utilitarian attitude scale, there are about 20 items respondents have to rate on a 7 point Likert scale.

I'm not expecting that everything loads the same, but I would expect that an item that measures the hedonic dimension in the 'before-question', would also do so in 'the after-question' and not that it starts loading on the utilitarian dimension...
Or should I just delete the item in that case in the 'after-question' ?

(it is a principal component factor analysis)

[Swank, Paul R]

You need to use SEM to test for invariance across time. If it is not there then comparing the measures across time is confounded with the change in factor structure.

Dr. Paul R. Swank,
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 kimberley2314
Sent: Thursday, April 28, 2011 5:00 PM
To: [hidden email]
Subject: Re: Before and after experiment using factor analysis [Sec: UNCLASSIFIED]

First of all, thanks for the answers!

The scale that I'm is using is a hedonic/utilitarian attitude scale, there
are about 20 items respondents have to rate on a 7 point Likert scale.

I'm not expecting that everything loads the same, but I would expect that an
item that measures the hedonic dimension in the 'before-question', would
also do so in 'the after-question' and not that it starts loading on the
utilitarian dimension...
Or should I just delete the item in that case in the 'after-question' ?

(it is a principal component factor analysis)--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Before-and-after-experiment-using-factor-analysis-tp4346460p4354129.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
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

[David Marso]

Do you or the instrument's authors have a 'theory/model' regarding which items should load on which of the 2 'factors' (If so perhaps consider using CFA rather than EFA)?
Are the shifting items loading strongly in each of the before/after 'factors'?
What do you mean by 'comparing'?  Are you attempting to address changes in the means of these 'factors'?
If so then CFA provides a relatively simple means of achieving this.
How many respondents do you have? 10-20 per variable are typically desired for PCA (more for CFA).
If you calculate 'factor scores' for each of the 2 'factors' for each time point what are the correlations between these scores across time points (T1F1,T2F1),(T1F1,T2F2),(T1F2,T2F2),(T1F2,T2F1)???
What are the reliabilities of 'scales' comprising of the salients for the various measures (unit weights).
I believe these are important issues to address.
---

kimberley2314 wrote

First of all, thanks for the answers!

The scale that I'm is using is a hedonic/utilitarian attitude scale, there are about 20 items respondents have to rate on a 7 point Likert scale.

I'm not expecting that everything loads the same, but I would expect that an item that measures the hedonic dimension in the 'before-question', would also do so in 'the after-question' and not that it starts loading on the utilitarian dimension...
Or should I just delete the item in that case in the 'after-question' ?

(it is a principal component factor analysis)