Many researchers have opted for using wavelet transformation for data modelling. The data is transformed using wavelets, and the decomposed coefficients ( approximation and detail coefficients) are individually fitted into distribution models (Gamma, Gaussian...etc). Why is that a better option, rather than trying to fit the original data. What statistical parameters are gained using this wavelet transformation? Which types of data are usually better modelled using the wavelet transform and then applying statistical techniques to them?
Thank you.