Hello,
I need some guidance with the following:
Can I do Principal Component Analysis (PCA) using a polychotic correlation matrix if I am dealing with unordered categorical data (multiple response) . I want to analyze if the answers selected in one variable impact the outcomes observed in the second. My understanding is that PCA for categorical data needs to use a polychoric correlation matrix instead of the Pearson matrix. However, I am not sure if my variables are acceptable for that analysis. I read that the variables need to be ordered. Do you think that multiple correspondence analysis is a better option? 

Any guidance is appreciated. Thanks.

Sample Data 
Variable 1-- all that apply
Do you like blue
0-no
1-yes
Do you like green
0-no
1-yes
Do you like red
0-no
1-yes
Variable 2--all that apply
Reason for color selection
1-I like the moon
2-I like the sun
3-I like the stars
4-I like the sky


#SPSSStatistics

Original Message:
Sent: 5/15/2021 10:50:00 PM
From: Rose
Subject: Principal Component Analysis

Hello,
I need some guidance with the following:
Can I do Principal Component Analysis (PCA) using a polychotic correlation matrix if I am dealing with unordered categorical data (multiple response) . I want to analyze if the answers selected in one variable impact the outcomes observed in the second. My understanding is that PCA for categorical data needs to use a polychoric correlation matrix instead of the Pearson matrix. However, I am not sure if my variables are acceptable for that analysis. I read that the variables need to be ordered. Do you think that multiple correspondence analysis is a better option? 

Any guidance is appreciated. Thanks.

Sample Data 
Variable 1-- all that apply
Do you like blue
0-no
1-yes
Do you like green
0-no
1-yes
Do you like red
0-no
1-yes
Variable 2--all that apply
Reason for color selection
1-I like the moon
2-I like the sun
3-I like the stars
4-I like the sky




------------------------------
Rose
------------------------------

#SPSSStatistics

Hi,

Thank you so much for your guidance. I will look into the CATREG and GENLIN suggestions.

Highly Appreciated,

Tania. 

------------------------------
Rose
------------------------------

Original Message:
Sent: Mon May 17, 2021 05:06 PM
From: Jon Peck
Subject: Principal Component Analysis

If all the variables are multiple nominal, 

"

• Multiple nominal. The only information in the observed variable that is preserved in the optimally scaled variable is the grouping of objects in categories. The order of the categories of the observed variable is not preserved. Category points will be in the centroid of the objects in the particular categories. Multiple indicates that different sets of quantifications are obtained for each dimension.

CATPCA and MULTIPLE CORRESPONDENCE will give the same results, but you don't have a dependent variable in those procedures.  You might want to consider CATREG, which has a dependent variable and can treat the regressors as nominal variables or GENLIN.

--

Jon K Peck
jkpeck@gmail.com

Original Message:
Sent: 5/15/2021 10:50:00 PM
From: Rose
Subject: Principal Component Analysis

Hello,
I need some guidance with the following:
Can I do Principal Component Analysis (PCA) using a polychotic correlation matrix if I am dealing with unordered categorical data (multiple response) . I want to analyze if the answers selected in one variable impact the outcomes observed in the second. My understanding is that PCA for categorical data needs to use a polychoric correlation matrix instead of the Pearson matrix. However, I am not sure if my variables are acceptable for that analysis. I read that the variables need to be ordered. Do you think that multiple correspondence analysis is a better option? 

Any guidance is appreciated. Thanks.

Sample Data 
Variable 1-- all that apply
Do you like blue
0-no
1-yes
Do you like green
0-no
1-yes
Do you like red
0-no
1-yes
Variable 2--all that apply
Reason for color selection
1-I like the moon
2-I like the sun
3-I like the stars
4-I like the sky




------------------------------
Rose
------------------------------

#SPSSStatistics