Some numerical experiments with recent offshore tide measurements have examined various parameters involved in tidal prediction by the response method: the number of prediction weights, their lead (and lag) times, and the treatment of radiational tides. The optimum number of weights depends directly on the length of record and inversely on noise level in a tidal band; more weights degrade the prediction and generate an artificial wiggliness in the admittance.
Zetler, Bernard D., and Walter H. Munk. 1975. "The optimum wiggliness of tidal admittances." Journal of Marine Research 33, (S). https://elischolar.library.yale.edu/journal_of_marine_research/1331