SPSSX Discussion

Factor analysis

Classic

List

Threaded

4 messages Options

Hong Wan

Factor analysis

Dear All

I am currently having difficulties in interpreting the results shown in rotated component matrix using principal components with Varimax rotation.

Some results provided by SPSS are listed as follows:

· Bartlett’s test of Sphericity (χ2 (276) = 543.180, p = .000)

· KMO (p = .000)

· Determinant of the R-matrix was 0.001

· Communalities of all the items were above 0.5

· % of the total variance is 59.873

Based on these results, is it possible for me to know whether it is appropriate to use factor analysis on my data?

If it is appropriate for me to use factor analysis, has anybody experienced the difficulties related to how to interpreting the factors? If so, could anybody give me some suggestions on how to how to deal with these?

Thank you very much in advance!

Best wishes

Yours sincerely,
Lily

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

SR Millis-3

Re: Factor analysis

I think that you need to provide more information:

--sample size

--the number of variables

--the scaling of the variables

--description of the variables

--description of your sample

~~~~~~~~~~~
Scott R Millis, PhD, ABPP, CStat, CSci
Professor & Director of Research
Dept of Physical Medicine & Rehabilitation
Dept of Emergency Medicine
Wayne State University School of Medicine
261 Mack Blvd
Detroit, MI 48201
Email: [hidden email]
Email: [hidden email]
Tel: 313-993-8085
Fax: 313-966-7682

--- On Sat, 5/8/10, Hong Wan <[hidden email]> wrote:

> From: Hong Wan <[hidden email]>
> Subject: Factor analysis
> To: [hidden email]
> Date: Saturday, May 8, 2010, 10:13 PM
> Dear All
>
> I am currently having difficulties in interpreting the
> results shown in rotated component matrix using principal
> components with Varimax rotation.
>
> Some results provided by SPSS are listed as follows:
>
> ·� � � � � � � �
> � � � � � Bartlett’s test of
> Sphericity (χ2 (276) = 543.180, p = .000)
>
> ·� � � � � � � �
> � � � � � KMO (p = .000)
>
> ·� � � � � � � �
> � � � � � Determinant of the
> R-matrix was 0.001
>
> ·� � � � � � � �
> � � � � � Communalities of all the
> items were above 0.5
>
> ·� � � � � � � �
> � � � � � % of the total variance
> is 59.873
>
> Based on these results, is it possible for me to know
> whether it is appropriate to use factor analysis on my
> data?
>
> If it is appropriate for me to use factor analysis, has
> anybody experienced the difficulties related to how to
> interpreting the factors? If so, could anybody give me some
> suggestions on how to how to deal with these?
>
> Thank you very much in advance!
>
> Best wishes
>
> Yours sincerely,
> Lily
>
> =====================
> 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

SR Millis-3

Re: Factor analysis

In reply to this post by Hong Wan

This additional information is needed to determine whether factor analysis is appropriate. For example, if you variables have ordinal scaling, standard FA may not be appropriate. Small sample size can also create problems. You also need to tell us whether you performed PCA or FA.

SR Millis
~~~~~~~~~~~
Scott R Millis, PhD, ABPP, CStat, CSci
Professor & Director of Research
Dept of Physical Medicine & Rehabilitation
Dept of Emergency Medicine
Wayne State University School of Medicine
261 Mack Blvd
Detroit, MI 48201
Email: [hidden email]
Email: [hidden email]
Tel: 313-993-8085
Fax: 313-966-7682

--- On Sun, 5/9/10, Hong Wan <[hidden email]> wrote:

> From: Hong Wan <[hidden email]>
> Subject: RE: Factor analysis
> To: "SR Millis" <[hidden email]>
> Date: Sunday, May 9, 2010, 10:34 AM
> Dear Prof. Millis
>
> May I ask why there are needs for further information? Is
> it possible for me to know it with further details?
>
> Thank you!
>
> Best wishes
>
> Hong Wan
>
> ________________________________________
> From: SR Millis [[hidden email]]
> Sent: 09 May 2010 15:28
> To: Hong Wan; SPSS
> Subject: Re: Factor analysis
>
> I think that you need to provide more information:
>
> --sample size
>
> --the number of variables
>
> --the scaling of the variables
>
> --description of the variables
>
> --description of your sample
>
>
> ~~~~~~~~~~~
> Scott R Millis, PhD, ABPP, CStat, CSci
> Professor & Director of Research
> Dept of Physical Medicine & Rehabilitation
> Dept of Emergency Medicine
> Wayne State University School of Medicine
> 261 Mack Blvd
> Detroit, MI 48201
> Email:� [hidden email]
> Email:� [hidden email]
> Tel: 313-993-8085
> Fax: 313-966-7682
>
>
> --- On Sat, 5/8/10, Hong Wan <[hidden email]>
> wrote:
>
> > From: Hong Wan <[hidden email]>
> > Subject: Factor analysis
> > To: [hidden email]
> > Date: Saturday, May 8, 2010, 10:13 PM
> > Dear All
> >
> > I am currently having difficulties in interpreting
> the
> > results shown in rotated component matrix using
> principal
> > components with Varimax rotation.
> >
> > Some results provided by SPSS are listed as follows:
> >
> > ·
> >� � � �
> � � � Bartlett’s test of
> > Sphericity (χ2 (276) = 543.180, p = .000)
> >
> > ·
> >� � � � � � � KMO (p =
> .000)
> >
> > ·
> >� � � �
> � � � Determinant of the
> > R-matrix was 0.001
> >
> > ·
> >� � � �
> � � � Communalities of all the
> > items were above 0.5
> >
> > ·
> >� � � � � � � % of the
> total variance
> > is 59.873
> >
> > Based on these results, is it possible for me to know
> > whether it is appropriate to use factor analysis on
> my
> > data?
> >
> > If it is appropriate for me to use factor analysis,
> has
> > anybody experienced the difficulties related to how
> to
> > interpreting the factors? If so, could anybody give me
> some
> > suggestions on how to how to deal with these?
> >
> > Thank you very much in advance!
> >
> > Best wishes
> >
> > Yours sincerely,
> > Lily
> >
> > =====================
> > 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
> >

Hector Maletta

Re: Factor analysis

In reply to this post by Hong Wan

If the determinant is exactly zero, no solution can be computed. If the
determinant is very small, like 0.0001 or 0.00001, the results may be
unstable, in the sense that a small change in some variable values may
produce large changes in results. This is so for any regression or
correlation problem, including factor analysis.

This is due to some variable/s being an almost exact linear function of one
or more other variables (for instance, a total score in a scale is usually
an exact linear function of the items in the scale; but suppose the values
of the items originally had decimals that were rounded, then the total score
(obtained with the unrounded values) may be ALMOST an exact linear function
of the rounded items.

You may try eliminating one variable or another (choose whichever ones you
deem less important, or more closely related, conceptually, to other
variables), and see whether the value of the determinant significantly
increases.
Otherwise, you may use your data as they are, with det=0.0001, but beware of
the instabilities. These instabilities increase as samples get smaller, and
are very large with relatively small samples (i.e. less than, say, 50 cases
per variable involved in the factor analysis).

Hector

-----Mensaje original-----
De: Hong Wan [mailto:[hidden email]]
Enviado el: Sunday, May 09, 2010 1:24 PM
Para: Hector Maletta
Asunto: RE: Factor analysis

Dear Hector

Thank you very much for your e-mail.

According to Field (2005), in factor analysis, 'multicollinearity can be
detected by looking at the determinant of the R-matrix, which should be
greater than 0.00001', I am not sure that I understand your explanation
fully. Are there any possibility for you to provide me further details?

I do appreciate your help!

Best wishes

Yours sincerely,
Lily

ps Field, A (2005) Discovering Sataistics unsing spss. London: SAGE.

________________________________________
From: Hector Maletta [[hidden email]]
Sent: 09 May 2010 05:05
To: Hong Wan
Subject: RE: Factor analysis

The determinant of the correlation matrix (0.001) is too close to zero,
indicating that at least one variable is an almost exact linear function of
(some of) the others (which is called colinearity). If such is the case, the
results may be quite unstable (a small change in one value of one variable
may considerably alter the results). All the rest seems OK.
On the other hand, getting results that are not easily interpretable is a
common occurrence in Factor Analysis. Revise your variables in the light of
your theory.

Hector

-----Mensaje original-----
De: SPSSX(r) Discussion [mailto:[hidden email]] En nombre de Hong
Wan
Enviado el: Saturday, May 08, 2010 11:13 PM
Para: [hidden email]
Asunto: Factor analysis

Dear All

I am currently having difficulties in interpreting the results shown in
rotated component matrix using principal components with Varimax rotation.

Some results provided by SPSS are listed as follows:

· Bartlett's test of Sphericity (χ2 (276) =
543.180, p = .000)

· KMO (p = .000)

· Determinant of the R-matrix was 0.001

· Communalities of all the items were above 0.5

· % of the total variance is 59.873

Based on these results, is it possible for me to know whether it is
appropriate to use factor analysis on my data?

If it is appropriate for me to use factor analysis, has anybody experienced
the difficulties related to how to interpreting the factors? If so, could
anybody give me some suggestions on how to how to deal with these?

Thank you very much in advance!

Best wishes

Yours sincerely,
Lily

=====================
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
Se certificσ que el correo entrante no contiene virus.
Comprobada por AVG - www.avg.es
Versiσn: 8.5.437 / Base de datos de virus: 271.1.1/2840 - Fecha de la
versiσn: 05/08/10 18:26:00
Se certificσ que el correo entrante no contiene virus.
Comprobada por AVG - www.avg.es
Versiσn: 8.5.437 / Base de datos de virus: 271.1.1/2840 - Fecha de la
versiσn: 05/08/10 18:26:00

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