Attitudinal Segmentation methodology advice

[Mark Webb-3]

Can group members comment on the pro & cons of these 2 methods for
segment on attitudinal statements.

When is it best use each method?
Do any have glaring strengths or weaknesses vs the other.

Method 1
Factor analysis and then chose the highest factor score [P factor analysis].
This essentially clusters variables.

Method 2
Factor analysis & then some cluster analysis based on the factor scores.
This essentially clusters respondents.

Regards

--
Mark Webb

[phone]
[phone]
Skype - webbmark
[hidden email]

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

[Art Kendall]

attitudes are usually measured by scales based on factor analysis of items.

The number of scales derived is very dependent on the actual items that
were used.
Many attitude domains have more than one dimension.

Strict factor scores often over capitalize on chance and unique variance
of items.  So it is conventional to retain as many factors as you can
get a good interpretation of using only items that load over some cutoff
e.g., |.45| on a factor and not over some other cutoff e.g., |.30| on
any other.
Items that have negative signs are reflected, i.e., 1 to 5 becomes 5 to 1.

Most methods for clustering cases assume that the variables that go in
ae fairly indepedent, so clustering of raw items is seldom done.  Scale
scores are usually more appropriate.

If you only have one attitude factor, clustering becomes much less
meaningful.


Art Kendall
Social Research Consultants



Mark Webb wrote:

> Can group members comment on the pro & cons of these 2 methods for
> segment on attitudinal statements.
>
> When is it best use each method?
> Do any have glaring strengths or weaknesses vs the other.
>
> Method 1
> Factor analysis and then chose the highest factor score [P factor
> analysis].
> This essentially clusters variables.
>
> Method 2
> Factor analysis & then some cluster analysis based on the factor scores.
> This essentially clusters respondents.
>
> Regards
>
> --
> Mark Webb
>
> [phone]
> [phone]
> Skype - webbmark
> [hidden email]
>
> =====================
> 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

[Zetu, Dan]

The method that you use really depends on what you want to accomplish.
If your goal is data reduction, then clustering of variables should be
used. If, on the other hand, you want to group your cases/respondents
into meaningful segments in terms of attitudes, then you would want to
cluster the cases. However, the two are not necessarily mutually
exclusive, you may want to cluster variables first to mitigate
multicollinearity, and then cluster the cases, as Art has mentioned
below.

-----------------------------
Dan Zetu
Analytical Consultant
R.L. Polk & Co.
248-728-7278
[hidden email]


-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Art Kendall
Sent: Friday, May 22, 2009 7:49 AM
To: [hidden email]
Subject: Re: Attitudinal Segmentation methodology advice

attitudes are usually measured by scales based on factor analysis of
items.

The number of scales derived is very dependent on the actual items that
were used.
Many attitude domains have more than one dimension.

Strict factor scores often over capitalize on chance and unique variance
of items.  So it is conventional to retain as many factors as you can
get a good interpretation of using only items that load over some cutoff
e.g., |.45| on a factor and not over some other cutoff e.g., |.30| on
any other.
Items that have negative signs are reflected, i.e., 1 to 5 becomes 5 to
1.

Most methods for clustering cases assume that the variables that go in
ae fairly indepedent, so clustering of raw items is seldom done.  Scale
scores are usually more appropriate.

If you only have one attitude factor, clustering becomes much less
meaningful.


Art Kendall
Social Research Consultants



Mark Webb wrote:

> Can group members comment on the pro & cons of these 2 methods for
> segment on attitudinal statements.
>
> When is it best use each method?
> Do any have glaring strengths or weaknesses vs the other.
>
> Method 1
> Factor analysis and then chose the highest factor score [P factor
> analysis].
> This essentially clusters variables.
>
> Method 2
> Factor analysis & then some cluster analysis based on the factor

scores.

> This essentially clusters respondents.
>
> Regards
>
> --
> Mark Webb
>
> [phone]
> [phone]
> Skype - webbmark
> [hidden email]
>
> =====================
> 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