See all my videos at www.tilestats.com/ 1. Introduction 2. Collinearity (01:07) 3. How PCR works (03:46) 4. Predict (08:30) 5. Extract components(08:50) 6. PCR vs PLS (13:17)
Thanks a lot. I have got a question: that's right, PCR is designed to resolve regression tasks. And what about classification? For instance, two classes are assigned as an output. The variables are reduced to the principal components, ok. Is it correct to use the obtained PCs for classification instead of original variables afterwards? Thanks for consultation!
Yes, that works. Have a look at this video about LDA where I compare PCA and LDA for classification: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-julEqA2ozcA.html I also have a video about PLS-DA ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-zqFZ0mEc74k.html
If I want to fit a linear regression model after pca, do I need to regress y on centered score or centered y on centered score? I have been confused about this question.
@@tilestats Not happening. I hope that you might be having some country specific criteria. will suggest you to kindly email me a separate link for making payment.
The scores of PC1 were rounded, which explains the difference. If you use the rounded scores, you will get what you say. Here are more exact values for PC1: 104.9293 107.7236 109.3397 110.5179 112.1340 114.9283
Thank you! I know PCA regression is an alternative of a multivariate regression to deal with multicollinearity. So, is PCA regression always better for prediction than a standard linear regression?