# whether to use normalization while performing PCA?

I have an excel file containing a table where I registered the frequency of three linguistic phenomena in 72 poems. Since the poems have different lengths I normalized the results dividing each value by the number of tokens of each poem and multiplying per 100. After that, I opened orange, added a new file, connected data table and selected the table. At this point, I connected PCA and with two components I have 97% explained variance but only if I deselect the "normalize variables" option and then I have a nice scatter plot.

My question is the following: when do I need to use the "normalize variables" option? Because in this case, I deselected it because I had already normalized the data. However, I am not sure if I made any mistakes.

• Concept of "normalize variables" is used by some authors without any conscious analysis. Deselecting the option is helping for this reason. PCA has no con nection with it. Further, you appear to follow some procedure that does not implement "normalization ". Jan 24, 2021 at 6:06