My professor wants us to match a number of ARMA time series with charts like these. Typically it would be like 5 series and then 5 corresponding pairs of such ACF/PACF charts. Sadly I have no real clue on how to interpret them, and I haven't found anything in the library yet. However, i suspect that there might be a number of rather simple rules like "if there is a spike at position x, then it's such and such process."Help much appreciated!CM

hint: go read Granger & Newboldregards,

I'm having trouble reading the picture, but it looks as if you have PACF charts on the left and ACF on the right. The first rule is to look at the simpler one.On the top line, the PACF is simpler. There is only interaction at lag one, every other lag just has the correlation implied by the one-lag effect. So (1,0,0) seems right.On the second line, the ACF is simpler. There are interactions at lags one and two, but none at other lags, not even the ones implied by the one and two lag effect. So (0,0,2) seems right.These are perfect charts for simple cases. Do you see the logic for these? If not, I can explain it If so, do you have a problem with messier charts and more complicated models?

I would also add that there is evidence of seasonality in the bottom 2 PACF charts. Hence, a Box-Jenkins SARIMA model would be appropriate in this situation. Check out The Analysis of Time Series by Chris Chatfield along with HP Scanjet User Manual. Both great books in their own right.