Time Series Book


Less is more. Stay pure. Stay poor.
As I mentioned before I had been treating repeated measures with multilevel models, given I don't typically have many panel measures. This and I don't think there is any moving averages to control for. Though, as I noted, I am slowing trying to pick up some time series knowledge. I am still a novice and the multivariate time series is beyond my current level, since I still need to master the univariate models.

What do you mean by ARDL in particular? Is this the process to determine how long a lag should be based on AR?

Oddly, I saw a printed chapter sitting in my closet on Saturday on forecasting using SAS, but it didn't say what book it was from, just Chapter 2? If Ii remember, I will try to look up the source tomorrow. You are lucky you couldn't open GG's links. I did and they were full of propaganda and evil. Just kidding, even worse for you (per your prior posts), most were VERY formula heavy.


No cake for spunky
ARDL is Autoregressive Distributed Lag. It assumes that lags of Y and predictors influence Y. It used to be among the most important multivariate approaches to time series although VARMA or VECM appear to have replaced it in that regard.

Matrices are evil and propaganda :( I have read time series books that are more matrices/formulas than words....

The forecasting book using SAS is probably by Dickey et el. Its written in 2003 and thus is badly dated as software (probably methods too). I avoid SAS software books more than five years old because the software changes so much.


No cake for spunky
Chapter 2 in "SAS for Forecasting Time Series" talks about forecasting with PROC ARIMA, the backshift notation and Yule Walker. It starts on page 29 and ends on p 48.