I am running PCA on a dataset with dimensions 1460(n) * 307(p). Upon running PCA, I obtain the below plot of variance explained and the number of components. I am trying to estimate the number of PCs to select for my training dataset.

Based on the plot below, It appears that around 75% of the variance is explained/captured by 50 PCs. So, selecting 50 PCs is a reasonable estimate.

How would you interpret the below plot and estimate # of PCs to select?


enter image description here

  • $\begingroup$ support.minitab.com/en-us/minitab/17/topic-library/… $\endgroup$ – SmallChess Mar 22 '17 at 3:11
  • $\begingroup$ You choose your desired percentage (80%, 90%). You use it to select your components. $\endgroup$ – SmallChess Mar 22 '17 at 3:12
  • $\begingroup$ @Student T yeah - the scree plot shows a steep curve until like ~20 PCs and then it's horizontal line from then on. $\endgroup$ – kms Mar 22 '17 at 3:15
  • $\begingroup$ Answers in this post go into detail on methods for appropriately selecting the number of components. The Valle paper referred to is a very good introduction to this topic: stats.stackexchange.com/questions/100143/… $\endgroup$ – Deathkill14 Mar 22 '17 at 6:42

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.