What is the advantages of lda over pca? Where they can be used?
Answers
Heya mate:)
PCA is a special case of factor analysis where factors are found using spectral decomposition of Eigen values. There are other rotations (later in the answer) that use factor analysis and not PCA.
But there are some inherent problems with PCA (tying back to your question here) that have caused a reduction in its usage. PCA or principal component analysis has three properties viz. maximizing the variance, orthogonality of basis and dimension reduction.
But due to the first point i.e. maximization of variance in the p-dimensional space of the data under a quadratic constraint, leads to a large number of variables loading in on the first principal component. Although this leads to effective dimension reduction, there is little more you can do with these PCs as the interpretation is lost.
Hope my answer helps you:)