I know this topic of scaling and normalizing variables for PCA has been posted on a lot, 1, 2, 3. However, I am performing PCA on coordinate data that is measured all on the same scale, i.e. (x,y) data. Would I have to standardize or normalize this data to account for the variance? Or is that not neccassary since the variables are measured on the same scale.

I tested an instnace within my data and found that when I do not scale or normalize the data, I explain 73.8% of the variance in the x direction and 26.1% of the variacne in the y direction. When I scale the data, 88.6% in the x and 11.4% in the y. Normalizing the data doesn't effect the results when there is no scaling because the coordiante data is the range from [0,100], which can be scaled to [0,1].

To be specific, I am performing PCA on coordinate data of individual players to find their area. So I have all the xy data of each player in a set. Before I run PCA on each player I would scale the entire dataset. Then I filtered to get the player, applied PCA, extracted the standard deviation, then used the elipsodial area equation to find the area.


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.