Basically the title asks my question. I have the results of a survey that was filled out by people from different countries. I have been asked to analyze the data using PCA and see what findings I can get on each country from my analysis. Now of course this means I have to aggregate my data, but should I aggregate the data before or after I standardize the data? From the results that I am getting it seems that aggregating first is the right way to go about it, but I am failing to understand why that would be the case.
-
$\begingroup$ what exaxtly is aggregated and what is your task (prediction?) $\endgroup$– PeterOct 20, 2020 at 17:58
1 Answer
Data should be standardized for calculating principal components. If data is standardized before aggregating by country, the resultant data after aggregation will no longer will have mean 0 and standard deviation 1, hence, the results of PCA will be erroneous. Therefore, data should be aggregated and then standardized in order to obtain correct principal components of the data.