CATPCA Questions

[cjbc]

Good morning,

When we choose CATPCA in Optimal Scaling in SPSS, I would like to know the importance of scale and weight in the analysis. For example, now i am working with variables with 6 categories each one. The purpose of CATPCA is to obtain a quantification of each categorie of the categorical variables to implement later PCA.

I have a problem because i am choosing variables that are correlated with the poverty, but when SPSS show the quantification starts with negatives values. I have read that the order of the categories are not relevant for the quantification. For example the categories are organized in descendent order. One of this variables is wall material of housing, and it is organized from the best to worst material of wall; the quantifications show negatives values for the best material and positives values for worst material. The theory shows that the results are inversely, because when a person have a better housing condition will obtain more value in the related categorie.

I want to know if CATPCA show quantifications starting of the assumption that the categories are organized from lowest to highest position. Adittionaly, it will be important to know if a negative quantification value is possible, becuase i have readed different papers about this topic and all agree that the values start from cero, and the scale is positive.

Thanks.

[Hector Maletta]

CATPCA would work with the initial values provided, and then gradually
arrive at optimal values. But the initial values of the categories (e.g. 1,
2, 3, ....) are increasing, and therefore the program would assume the
underlying quantitative variable has the same direction. If your categories
increase as "poverty" decreases, then those variables indicate
"non-poverty", and the values would be calculated accordingly. You may want
to recodify the observed variables assigning numeric values in the direction
wanted (e.g. 6, 5, 4, 3, 2 and 1). Otherwise the results would be "upside
down". Also, it is better that this is so for all variables (in a recent
analysis of mine, I had several indicators with values corresponding to
increasing levels of wellbeing, such as housing materials ordered from worst
to best, but one indicator was "persons per room" which is decreasing in
wellbeing, since more persons per room mean more overcrowding; I decided to
use "rooms per person" instead).
CATPCA accepts any variable, but starts with the order given in the initial
values. If all variables go from best to worst, it would be too much to ask
for the procedure to reverse the order of all variables, especially because
the correlate (poverty) is not used in the CATPCA calculation.

Hector

-----Mensaje original-----
De: SPSSX(r) Discussion [mailto:[hidden email]] En nombre de cjbc
Enviado el: Wednesday, July 28, 2010 11:55 AM
Para: [hidden email]
Asunto: CATPCA Questions

Good morning,

When we choose CATPCA in Optimal Scaling in SPSS, I would like to know the
importance of scale and weight in the analysis. For example, now i am
working with variables with 6 categories each one. The purpose of CATPCA is
to obtain a quantification of each categorie of the categorical variables to
implement later PCA.

I have a problem because i am choosing variables that are correlated with
the poverty, but when SPSS show the quantification starts with negatives
values. I have read that the order of the categories are not relevant for
the quantification. For example the categories are organized in descendent
order. One of this variables is wall material of housing, and it is
organized from the best to worst material of wall; the quantifications show
negatives values for the best material and positives values for worst
material. The theory shows that the results are inversely, because when a
person have a better housing condition will obtain more value in the related
categorie.

I want to know if CATPCA show quantifications starting of the assumption
that the categories are organized from lowest to highest position.
Adittionaly, it will be important to know if a negative quantification value
is possible, becuase i have readed different papers about this topic and all
agree that the values start from cero, and the scale is positive.

Thanks.
--
View this message in context:
http://spssx-discussion.1045642.n5.nabble.com/CATPCA-Questions-tp2256689p225
6689.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
Se certificó que el correo entrante no contiene virus.
Comprobada por AVG - www.avg.es
Versión: 8.5.441 / Base de datos de virus: 271.1.1/3027 - Fecha de la
versión: 07/28/10 08:37:00

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