I'm struggling with this problem; on p. 125-6 of the Cryer text, they imply that the first difference of the log of gas prices is a moving average model, so the original data is an IMA(1,1) model, because only the sample autocorrelation at lag 1 is significant. However, in the sample Excel file posted on the discussion board "TimeSeries_SampleProject.xls", tab "Corrg First Diffcs Real Int Rat", the box says that an autoregressive model is implied from the same logic. NEAS or fellow actuaries, can you help me with this?
[NEAS: Cryer and Chan say: "After differencing, a moving average model of order 1 seems appropriate." They infer this from the correlogram for this time series; it is not a general statement. If the original time series has an exponential trend, the first differences of the logarithms is an autoregressive process.]