The example is as follows:

A bunch of doctors were asked to score a list of desirable characteristics of sales representatives. The questions were like: "in-depth knowledge about his/her product", "respectful of my time" etc. The goal is to understand if there are groups of doctors that value one type of traits more than the other type of traits and maybe target those doctors differently

Dataset generated: $\mathbf{D}$ of dimension $\mathbf{d} \mathbf{x} \mathbf{q}$ where $\mathbf{d}$ is the number of doctors and $\mathbf{q}$ is the number of questions.

Approach: A PCA is run and top 3 principal components are extracted. The pic is a bit blurry but, in the lecture I have referenced it is represented as follows: enter image description here


  1. The lecture mentions: These are not exactly the output of the PCA but once you have identified the top 3 principal components you run a postprocessing step in which you rotate the 3 principal components to ensure they have high values in some group of questions and low values in some other group of questions. This way the principal components are interpretable. How is this rotation of principal components achieved?
  2. How does the above post-processing help make the principal component analysis more interpretable? Is it because the doctors with higher projections on the rotated principal component 1 value medical knowledge more?
  3. Why is this dataset in the image represented like this? It gives the impression that the rows represent something. Is there any way to interpret the rows in the image?
  4. Any idea where this example is taken from? This seems like a good example of learning practical ways to use PCA. If not, could you reference some other similar practical example where PCA is used to understand the high dimensional data?

1 Answer 1


I find it useful to decode PCs using stepwise regression. Examples are here and here.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.