Multiple loadings for a variable in Factor analysis

[Naru]

Hi

I am carrying out factor analysis for my research data using SPSS. I have about 200 cases.

Some of the variables have multiple loadings to different factors in the rotated component matrix results.
I have given a cut off of 0.3 for "Coefficient display format - Suppress absolute values less than" option in SPSS.

There is one variable where the loading is 0.3 against Factor 1 and 0.61 against Factor-3.

I would want to classify this variable under Factor-1 rather than Factor-3.
I have read materials about factor analysis. None of them talk about moving a variable from one factor to another factor based on valid arguments.

Is that an acceptable practice ?

[Kathryn Gardner]

Hi,
 
I haven't read any guidelines as such, but i've seen other researchers move items. However, you could probably best argue for moving an item if the item has similarly sized loadings on the two factors e.g., .33 factor 1 .37 factor 2. If your item is theoretically more associated with factor 1, you have a reasonably good case for moving it. However, if an item has a larger loading on another factor as in your case, I think that moving it becomes more complicated, because the item is sharing more variance with the factor it loads highest on. If you moved it, this might then increase the correlations amongst the two factors since fsay actor 2 has an item that clearly (statistically) belongs to factor 1. If your item is loading highly on an unexpected factor, there should be some theoretical reason i.e., the item is not tapping the construct you thought it was.
 
I may be wrong on this though, and other people may be able to offer better advice.
 
Good luck
Kathryn> Date: Sat, 18 Aug 2007 17:00:11 -0700> From: [hidden email]> Subject: Multiple loadings for a variable in Factor analysis> To: [hidden email]> > Hi> > I am carrying out factor analysis for my research data using SPSS. I have> about 200 cases.> > Some of the variables have multiple loadings to different factors in the> rotated component matrix results.> I have given a cut off of 0.3 for "Coefficient display format - Suppress> absolute values less than" option in SPSS.> > There is one variable where the loading is 0.3 against Factor 1 and 0.61> against Factor-3.> > I would want to classify this variable under Factor-1 rather than Factor-3.> I have read materials about factor analysis. None of them talk about moving> a variable from one factor to another factor based on valid arguments.> > Is that an acceptable practice ?> --> View this message in context: http://www.nabble.com/Multiple-loadings-for-a-variable-in-Factor-analysis-tf4291887.html#a12218125> Sent from the SPSSX Discussion mailing list archive at Nabble.com.
_________________________________________________________________
Feel like a local wherever you go with BackOfMyHand.com
http://www.backofmyhand.com

[Noble, Lyndsay]

Hi,
Whether it is "acceptable" or not, I've been known to do it.  I always double
check the appropriateness by running internal reliability on my new factors.
Make sure anything that needs to be is reverse codes before running the
reliability or your alphas will be horrible.

Lyndsay


-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Naru
Sent: Saturday, August 18, 2007 5:00 PM
To: [hidden email]
Subject: Multiple loadings for a variable in Factor analysis

Hi

I am carrying out factor analysis for my research data using SPSS. I have
about 200 cases.

Some of the variables have multiple loadings to different factors in the
rotated component matrix results.
I have given a cut off of 0.3 for "Coefficient display format - Suppress
absolute values less than" option in SPSS.

There is one variable where the loading is 0.3 against Factor 1 and 0.61
against Factor-3.

I would want to classify this variable under Factor-1 rather than Factor-3.
I have read materials about factor analysis. None of them talk about moving
a variable from one factor to another factor based on valid arguments.

Is that an acceptable practice ?
--
View this message in context:
http://www.nabble.com/Multiple-loadings-for-a-variable-in-Factor-analysis-tf4
291887.html#a12218125
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

[Henrik Lolle]

Hi,

I aggree with Kathryn, but will further mention that if you have
access to AMOS you could do a confirmative factor analysis. You could
also, alternatively, take the variable out of the analysis and not
consider it for calculation of any factor. You could argue that the
variable is taken our because it loads highly on more than one factor
which is quite normal procedure.

Best,
Henrik
--
Henrik Lolle
Department of Economics, Politics and Public Administration
Aalborg University
Fibigerstraede 1
9200 Aalborg East
Phone: [phone]



Quoting Kathryn Gardner <[hidden email]>:

> Hi,
>
> I haven't read any guidelines as such, but i've seen other
> researchers move items. However, you could probably best argue for
> moving an item if the item has similarly sized loadings on the two
> factors e.g., .33 factor 1 .37 factor 2. If your item is
> theoretically more associated with factor 1, you have a reasonably
> good case for moving it. However, if an item has a larger loading on
>  another factor as in your case, I think that moving it becomes more
>  complicated, because the item is sharing more variance with the
> factor it loads highest on. If you moved it, this might then
> increase the correlations amongst the two factors since fsay actor 2
>  has an item that clearly (statistically) belongs to factor 1. If
> your item is loading highly on an unexpected factor, there should be
>  some theoretical reason i.e., the item is not tapping the construct
>  you thought it was.
>
> I may be wrong on this though, and other people may be able to offer
>  better advice.
>
> Good luck
> Kathryn> Date: Sat, 18 Aug 2007 17:00:11 -0700> From:
> [hidden email]> Subject: Multiple loadings for a
> variable in Factor analysis> To: [hidden email]> > Hi> > I
>  am carrying out factor analysis for my research data using SPSS. I
> have> about 200 cases.> > Some of the variables have multiple
> loadings to different factors in the> rotated component matrix
> results.> I have given a cut off of 0.3 for "Coefficient display
> format - Suppress> absolute values less than" option in SPSS.> >
> There is one variable where the loading is 0.3 against Factor 1 and
> 0.61> against Factor-3.> > I would want to classify this variable
> under Factor-1 rather than Factor-3.> I have read materials about
> factor analysis. None of them talk about moving> a variable from one
>  factor to another factor based on valid arguments.> > Is that an
> acceptable practice ?> --> View this message in context:
> http://www.nabble.com/Multiple-loadings-for-a-variable-in-Factor-analysis-tf4291887.html#a12218125> Sent from the SPSSX Discussion mailing list archive at
> Nabble.com.
> _________________________________________________________________
> Feel like a local wherever you go with BackOfMyHand.com
> http://www.backofmyhand.com

[Art Kendall-2]

With more than 35 years experience with factor analysis I have never
seen that done.
I would *not* do that.  I can think of no context in which that would be
appropriate.
What kind of argument would be used to do this?

Do you intend to interpret the dimensions?

Are you trying to verify that items can be used in scales?


Art Kendall
Social Research Consultants




Naru wrote:

> Hi
>
> I am carrying out factor analysis for my research data using SPSS. I have
> about 200 cases.
>
> Some of the variables have multiple loadings to different factors in the
> rotated component matrix results.
> I have given a cut off of 0.3 for "Coefficient display format - Suppress
> absolute values less than" option in SPSS.
>
> There is one variable where the loading is 0.3 against Factor 1 and 0.61
> against Factor-3.
>
> I would want to classify this variable under Factor-1 rather than Factor-3.
> I have read materials about factor analysis. None of them talk about moving
> a variable from one factor to another factor based on valid arguments.
>
> Is that an acceptable practice ?
> --
> View this message in context: http://www.nabble.com/Multiple-loadings-for-a-variable-in-Factor-analysis-tf4291887.html#a12218125
> Sent from the SPSSX Discussion mailing list archive at Nabble.com.
>
>
>