Hello,
I have calculated principal components for hundreds of respondents with an average of ten variables using Proc Factor. I was trying to use these individual principal components to compute single group measure for each component.
For instance, I have got 10 schools and 20 students per school and I got two principal components for the 10 variables.
How can I use these two components for each school?
Can I sum up comp1 to get one school measure for component one and take the average?
Do the same thing for comp2 and get one measure for each schools?
Component A = (sum of all comp1 for each student in school X) / (Total students in school X)
Component B = (sum of all comp2 for each student in school X) / (Total students in school X)
Regards
Teketo
How can I use these two components for each school?
That's really up to you. What would you like to learn from the data? You haven't told us.
Principal components are often used to help visualize data in lower dimensional space so you can detect patterns or clustering. Is that what you want?
Tell us what you want to learn from this data.
Can I sum up comp1 to get one school measure for component one and take the average?
Do the same thing for comp2 and get one measure for each schools?
Component A = (sum of all comp1 for each student in school X) / (Total students in school X)
Component B = (sum of all comp2 for each student in school X) / (Total students in school X)
Yes you can do this. However, taken exactly as you have described the question of comparing school averages, this is a situation where you would use MANOVA, rather than Principal Components.
So again, we need to know what you want to learn from this data.
You could, I'm assuming the 'groups' are predetermined?
I'm assuming you would graph it first to see if the distribution made sense between your groups? I think you're trying to SCORE your data. PROC SCORE has an example of how to get the principle component scores for your data, a fully worked example.
@Teketo wrote:
Hello,
I have calculated principal components for hundreds of respondents with an average of ten variables using Proc Factor. I was trying to use these individual principal components to compute single group measure for each component.
For instance, I have got 10 schools and 20 students per school and I got two principal components for the 10 variables.
How can I use these two components for each school?
Can I sum up comp1 to get one school measure for component one and take the average?
Do the same thing for comp2 and get one measure for each schools?
Component A = (sum of all comp1 for each student in school X) / (Total students in school X)
Component B = (sum of all comp2 for each student in school X) / (Total students in school X)
Regards
Teketo
@Reeza wrote:
You could, I'm assuming the 'groups' are predetermined?
I'm assuming you would graph it first to see if the distribution made sense between your groups? I think you're trying to SCORE your data. PROC SCORE has an example of how to get the principle component scores for your data, a fully worked example.
PROC SCORE is not really needed here, one of the output data sets from PROC PRINCOMP will contain the PCA scores. The only time PROC SCORE would be useful is to compute scores on observations that were not in the original data set.
How can I use these two components for each school?
That's really up to you. What would you like to learn from the data? You haven't told us.
Principal components are often used to help visualize data in lower dimensional space so you can detect patterns or clustering. Is that what you want?
Tell us what you want to learn from this data.
Can I sum up comp1 to get one school measure for component one and take the average?
Do the same thing for comp2 and get one measure for each schools?
Component A = (sum of all comp1 for each student in school X) / (Total students in school X)
Component B = (sum of all comp2 for each student in school X) / (Total students in school X)
Yes you can do this. However, taken exactly as you have described the question of comparing school averages, this is a situation where you would use MANOVA, rather than Principal Components.
So again, we need to know what you want to learn from this data.
Are you ready for the spotlight? We're accepting content ideas for SAS Innovate 2025 to be held May 6-9 in Orlando, FL. The call is open until September 25. Read more here about why you should contribute and what is in it for you!
ANOVA, or Analysis Of Variance, is used to compare the averages or means of two or more populations to better understand how they differ. Watch this tutorial for more.
Find more tutorials on the SAS Users YouTube channel.