Multimedialne seminarium z ekono- i socjofizyki

We can inexpensively model probability distribution as a polynomial: using orthonormal basis, the estimated coefficient of a function is just average of this function over the sample. Each such coefficient has a concrete cumulant-like interpretation, for joint distribution they describe statistical dependencies, we have some control of their accuracy. We can also model their time evolution and evolution of the density. Among others, I will talk about their application for time series analysis - for example modeling joint distribution of a few neighboring values to predict probability distribution of the next one based on a few previous values on the example of DJIA and yield curve parameters.