Can factors identified using factor analysis on a smaller dataset be used on larger dataset?

[Naila Baig-Ansari]

I used exploratory factor analysis to find 3 factors describing eating
patterns in healthy persons in my dataset.  I also generated factor
scores and examined the relationship of the scores obtained to a number
of demographic variables such as age, ethnicity, socioeconomic status,
etc.

 I would now like to use the 3 factors identified to my entire dataset
that includes healthy and unhealthy persons and examine relationships to
demographic variables.  How do I do this in SPSS?  Is there any way of
generating factor scores for the remaining dataset based on the factors
identified through the exploratory factor analysis?

Thanks for any insight
Naila

[statisticsdoc]

Stephen Brand
www.statisticsdoc.com

Naila,

You can compute factor scores using COMPUTE statements.  When you perform the factor analysis in the original sample, include /PRINT FSCORE on the command.  This option will generate weights that can be used to compute factor scores in the new sample.   Look for a matrix titled "Factor Score Coefficient Matrix".  Your factor analysis syntax should also SAVE factor scores (even if you are not going to use them again).  The factor score coefficients are computed using the method that you requested in the SAVE option.

To compute factor scores in the new sample, take the following steps.  Standardize all of the items.  For factor one, take the standardized score for each item and multiply it by its coefficient for factor one, and add the weighted scores.  Repeat this procedure for Factors Two and Three.
Now you have factor scores for the new sample using the scoring procedure from the old sample.

An alternative course of action might be to compute three composite variables by adding together the items that load highly and distinctively on each of the factors, and to weight the items equally.  This method is clearly not as precise, but it may be more robust across samples.

HTH,

Stephen Brand

---- Naila Baig-Ansari <[hidden email]> wrote:

> I used exploratory factor analysis to find 3 factors describing eating
> patterns in healthy persons in my dataset.  I also generated factor
> scores and examined the relationship of the scores obtained to a number
> of demographic variables such as age, ethnicity, socioeconomic status,
> etc.
>
>  I would now like to use the 3 factors identified to my entire dataset
> that includes healthy and unhealthy persons and examine relationships to
> demographic variables.  How do I do this in SPSS?  Is there any way of
> generating factor scores for the remaining dataset based on the factors
> identified through the exploratory factor analysis?
>
> Thanks for any insight
> Naila

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

[Swank, Paul R]

I would like to install a note of caution here. While it is perfectly
possible to compute factor scores for a new sample based on another the
solution in another sample, whether it is a good idea is open to
interpretation. The meaningfullness of the factor scores on the second
sample really depends on wether the factor structure is accurate for
that sample. A more appropriate tack may be to check for factorial
invariance across samples to ensure the structure is not different.

Paul R. Swank, Ph.D.
Professor, Developmental Pediatrics
Director of Research, Center for Improving the Readiness of Children for
Learning and Education (C.I.R.C.L.E.)
Medical School
UT Health Science Center at Houston

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Statisticsdoc
Sent: Thursday, June 29, 2006 10:16 AM
To: [hidden email]
Subject: Re: Can factors identified using factor analysis on a smaller
dataset be used on larger dataset?

Stephen Brand
www.statisticsdoc.com

Naila,

You can compute factor scores using COMPUTE statements.  When you
perform the factor analysis in the original sample, include /PRINT
FSCORE on the command.  This option will generate weights that can be
used to compute factor scores in the new sample.   Look for a matrix
titled "Factor Score Coefficient Matrix".  Your factor analysis syntax
should also SAVE factor scores (even if you are not going to use them
again).  The factor score coefficients are computed using the method
that you requested in the SAVE option.

To compute factor scores in the new sample, take the following steps.
Standardize all of the items.  For factor one, take the standardized
score for each item and multiply it by its coefficient for factor one,
and add the weighted scores.  Repeat this procedure for Factors Two and
Three.
Now you have factor scores for the new sample using the scoring
procedure from the old sample.

An alternative course of action might be to compute three composite
variables by adding together the items that load highly and
distinctively on each of the factors, and to weight the items equally.
This method is clearly not as precise, but it may be more robust across
samples.

HTH,

Stephen Brand

---- Naila Baig-Ansari <[hidden email]> wrote:
> I used exploratory factor analysis to find 3 factors describing eating

> patterns in healthy persons in my dataset.  I also generated factor
> scores and examined the relationship of the scores obtained to a
> number of demographic variables such as age, ethnicity, socioeconomic
> status, etc.
>
>  I would now like to use the 3 factors identified to my entire dataset

> that includes healthy and unhealthy persons and examine relationships
> to demographic variables.  How do I do this in SPSS?  Is there any way

> of generating factor scores for the remaining dataset based on the
> factors identified through the exploratory factor analysis?
>
> Thanks for any insight
> Naila

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

[statisticsdoc]

Stephen Brand
www.statisticsdoc.com

Nalia,

Professor Swank raises a good point here, leading to a few followup questions.

How large is your sample?  Do you have enough cases to carry out a confirmatory factor analysis in the well and sick group (i.e. looking at the structures separately in each group)?

Is there a compelling reason to factor analyze the scores in the well group, and then extend them to the larger well  plus sick group?   Would it make sense in view of your research questions to factor analyze the entire sample and make demographic comparisons on the overall factor structure for the healthy group and and for the sick group?

HTH,

Stephen Brand

---- "Swank wrote:

> I would like to install a note of caution here. While it is perfectly
> possible to compute factor scores for a new sample based on another the
> solution in another sample, whether it is a good idea is open to
> interpretation. The meaningfullness of the factor scores on the second
> sample really depends on wether the factor structure is accurate for
> that sample. A more appropriate tack may be to check for factorial
> invariance across samples to ensure the structure is not different.
>
> Paul R. Swank, Ph.D.
> Professor, Developmental Pediatrics
> Director of Research, Center for Improving the Readiness of Children for
> Learning and Education (C.I.R.C.L.E.)
> Medical School
> UT Health Science Center at Houston
>
> -----Original Message-----
> From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
> Statisticsdoc
> Sent: Thursday, June 29, 2006 10:16 AM
> To: [hidden email]
> Subject: Re: Can factors identified using factor analysis on a smaller
> dataset be used on larger dataset?
>
> Stephen Brand
> www.statisticsdoc.com
>
> Naila,
>
> You can compute factor scores using COMPUTE statements.  When you
> perform the factor analysis in the original sample, include /PRINT
> FSCORE on the command.  This option will generate weights that can be
> used to compute factor scores in the new sample.   Look for a matrix
> titled "Factor Score Coefficient Matrix".  Your factor analysis syntax
> should also SAVE factor scores (even if you are not going to use them
> again).  The factor score coefficients are computed using the method
> that you requested in the SAVE option.
>
> To compute factor scores in the new sample, take the following steps.
> Standardize all of the items.  For factor one, take the standardized
> score for each item and multiply it by its coefficient for factor one,
> and add the weighted scores.  Repeat this procedure for Factors Two and
> Three.
> Now you have factor scores for the new sample using the scoring
> procedure from the old sample.
>
> An alternative course of action might be to compute three composite
> variables by adding together the items that load highly and
> distinctively on each of the factors, and to weight the items equally.
> This method is clearly not as precise, but it may be more robust across
> samples.
>
> HTH,
>
> Stephen Brand
>
> ---- Naila Baig-Ansari <[hidden email]> wrote:
> > I used exploratory factor analysis to find 3 factors describing eating
>
> > patterns in healthy persons in my dataset.  I also generated factor
> > scores and examined the relationship of the scores obtained to a
> > number of demographic variables such as age, ethnicity, socioeconomic
> > status, etc.
> >
> >  I would now like to use the 3 factors identified to my entire dataset
>
> > that includes healthy and unhealthy persons and examine relationships
> > to demographic variables.  How do I do this in SPSS?  Is there any way
>
> > of generating factor scores for the remaining dataset based on the
> > factors identified through the exploratory factor analysis?
> >
> > Thanks for any insight
> > Naila
>
> --
> For personalized and experienced consulting in statistics and research
> design, visit www.statisticsdoc.com

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