help
help > RE: PCA decomposition of behavioural scores
Aug 12, 2017 01:08 AM | Alfonso Nieto-Castanon - Boston University
RE: PCA decomposition of behavioural scores
Hi Jenna,
Sorry for the late reply. Each of your two factors/components represents some combination of your original variables (6 subscales of the Repetitive Behaviour Scale). In particular each factor is a simple linear combination of the 6 subscale scores, and the weights associated with this linear combination can be obtained in several ways, perhaps one of the simplest is to use:
[Q,D,W]=svd(x);
in your original script when computing the factor/component scores. When doing this W(:,1) will be a normalized vector of 6 numbers characterizing the relative weights associated with each of your 6 subscales when defining the first factor/component, W(:,2) will be another normalized vector of 6 numbers now characterizing the second factor/component, etc. This weights can also be interpreted as indicative of the level of correlation between each of your factor/component scores and each of your original 6 subscale values. You may want to search for "principal component analysis interpretation" online to find a lot of examples of how one typically goes about interpreting these factors/components (or if you prefer, a good book/reference for the math behind this is Mardia's "Multivariate Analyses" book, and a good reference for the more practical side of things is Steven's "Applied Multivariate Statistics for the Social Sciences").
Hope this helps
Alfonso
Originally posted by Jenna Traynor:
Sorry for the late reply. Each of your two factors/components represents some combination of your original variables (6 subscales of the Repetitive Behaviour Scale). In particular each factor is a simple linear combination of the 6 subscale scores, and the weights associated with this linear combination can be obtained in several ways, perhaps one of the simplest is to use:
[Q,D,W]=svd(x);
in your original script when computing the factor/component scores. When doing this W(:,1) will be a normalized vector of 6 numbers characterizing the relative weights associated with each of your 6 subscales when defining the first factor/component, W(:,2) will be another normalized vector of 6 numbers now characterizing the second factor/component, etc. This weights can also be interpreted as indicative of the level of correlation between each of your factor/component scores and each of your original 6 subscale values. You may want to search for "principal component analysis interpretation" online to find a lot of examples of how one typically goes about interpreting these factors/components (or if you prefer, a good book/reference for the math behind this is Mardia's "Multivariate Analyses" book, and a good reference for the more practical side of things is Steven's "Applied Multivariate Statistics for the Social Sciences").
Hope this helps
Alfonso
Originally posted by Jenna Traynor:
Hi Alfonso,
I am sorry to keep bothering you about this, but I am having a lot of trouble finding the answer to my question elsewhere. I am completely done the analysis and just need to know what my component represents. Is it an amalgamation of scores from the original scale or does it represent something that I can identify from the original scale? (i.e., a specific score?)
If there is any way you can let me know how to go about finding this information or point me to a resource I would really appreciate it.
Thank you as always,
Jenna
I am sorry to keep bothering you about this, but I am having a lot of trouble finding the answer to my question elsewhere. I am completely done the analysis and just need to know what my component represents. Is it an amalgamation of scores from the original scale or does it represent something that I can identify from the original scale? (i.e., a specific score?)
If there is any way you can let me know how to go about finding this information or point me to a resource I would really appreciate it.
Thank you as always,
Jenna
Threaded View
Title | Author | Date |
---|---|---|
Jenna Traynor | Jul 26, 2017 | |
Jenna Traynor | Aug 11, 2017 | |
Alfonso Nieto-Castanon | Aug 12, 2017 | |
Jenna Traynor | Aug 13, 2017 | |