I have been running the KDE methods on my data of 26 variables and 500K observations. I'am able to use the densities to tag the outlier or potential outliers that I'am looking for. Although my final aim is to find the variables which if not used for the density estimation would lead that observation to be a part of normal population? Can you suggest some solution?
I have been running the KDE methods on my data of 26 variables and 500K observations. I'am able to use the densities to tag the outlier or potential outliers that I'am looking for. Although my final aim is to find the variables which if not used for the density estimation would lead that observation to be a part of normal population? Can you suggest some solution?