Skip to content

QnA.md

Latest commit

 

History

History
18 lines (9 loc) · 825 Bytes

QnA.md

File metadata and controls

18 lines (9 loc) · 825 Bytes

QnAs

  1. Can you compare variances of two variables?

     Nope. Only after normalisation.

  2. How many Principal Components should be there in PCA?

     = Number of variables. That's when all the PCs will explain all the variance in the data.

  3. What is the difference between Global Structure and Local Structure preservation during Dimensionality Reduction?

  4. t-Stats are basically necessary when sample size $n$ is <30. But in today's world, that is rarely a thing. So why is t-Stats still important!

  5. If association/correlation is sufficient for prediction, can we use spurious correlations for predictions?

Correlations have predictive power, when we assume that this correlation continues to exist (in future or on the test set we are predicting). A spurious correlation may not hold this prediction.