A GARCH(1, 1) + AR(1) model, where y(i) = c + phi * y(i - 1) + eta(i), and h(i), the variance of eta(i), is given by h(i) = omega + alpha * eta(i) ** 2 + beta * h(i - 1) ** 2
Models time dependent effects in a time series.
ARIMA models allow modeling timeseries as a function of prior values of the series (i.
Fits an Exponentially Weight Moving Average model (EWMA) (aka.
This model is basically a regression with ARIMA error structure see https://onlinecourses.science.psu.edu/stat510/node/53 https://www.otexts.org/fpp/9/1 http://robjhyndman.com/talks/RevolutionR/11-Dynamic-Regression.pdf The basic idea is that for usual regression models Y = B*X + e e should be IID ~ N(0,sigma2), but in time series problems, e tends to have time series characteristics.