Factor Analysis by Principal Component Method

Question

Armando MAROZZI 2020-3-21

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/512138-factor-analysis-by-principal-component-method

回答： Dinesh Yadav 2020-3-26

I am trying to replicate a paper that implements Factor Analysis by Principal Component Method. I have been reading a lot in Matlab but all the examples I see use MLE as an estimation method. Hence, my question is how can I change the estimation method to Principal Component Method in Factor Analysis? I just need unrotated factors at the moment to become familiar with Matlab.

Let's take this example:

load stocks
[Loadings,specificVar,T,stats] = factoran(stocks,3,'rotate','none'); %estimated by MLE

Is it possible to change the estimation strategy in that function?

I also found this that seems to be the only solution (at least to my limited knowledge)

function anfactpcwod(X)

Any advice?

2 个评论
显示无隐藏无

Jeff Miller 2020-3-22

Maybe you can get what you want by using the pca function instead of factoran.

I could be wrong, but I don't think MLE and PCA are alternatives in the sense that your question implies. MLE is an estimation criterion, like least squares. PCA and factor analysis differ with respect to the nature of the structure being estimated (by whatever criterion).

Armando MAROZZI 2020-3-22

PC is also an estimation method for Factor Analysis. How Principal Component differs from FA estimated with PC method, if any difference exists, I have no idea. I'll try "pca" and see whether the results differ. Thanks a lot!

请先登录，再进行评论。

请先登录，再回答此问题。