The choice of optimal calculation method for frequency-dependent polarizabilities (FDPs) is important for many applications. We compare some density functional theory (DFT) and ab initio methods with the quantum defect theory (QDT). Since QDT incorporates the experimental energy spectrum, it should yield more reliable results for FDP near the resonances at the excited states. We show that the DFT and ab initio methods which demonstrate the best accuracy for the static polarizabilities, have poor accuracy of the excited states energies. Such a behavior is a source of some discrepancies between these methods and QDT technique.