ECON 174 Lecture Notes - Lecture 18: Time Series, White Noise
2 Mar 2020
School
Department
Course
Professor
Document Summary
Can be used to forecast yt one period ahead. Cannot be used for longer forecast horizon. Cannot account for potential feedback effects from y to x. Income affects consumption, but consumption also affects income. Want to treat both variables as endogenous and predict jointly. Ensure all time series are (trend) stationary before including in a var. Before estimating a var, we need to select the number of time series (k). the number of lags included in the model (p) The number of parameters to estimate is then given by k + pk2. Can be demanding for short time series. Start with number of lags (p) that minimize the bic (sc). Estimate model, check if residuals are white noise. Try adding one additional lag to model if residuals fail test. We have data up to time t. i derive coefficient estimates using least square. Plug in coefficient estimates to derive 1-period ahead forecast:
