I'm doing a K-Means model for first time, thus very low experience. I read that it is not bad to discard variables through some PCA analysis. After standardizing the data, the loadings (weights) for two variables in the first component are very similar (0.52 and 0.53). These variables have 0.91 of correlation. In subsequent components both variables have weights close to zero.
Should one of them be removed? The total number of variables is 14.
Is there some threshold for the correlation to decide if its ok to remove a variable? Maybe I'm asking a 'it depends question', however further guidance will be appreciated.
