Here are my problems:
I am trying to model mens pole vault world records with an ARIMA model. When I graph the data it is linearly positive (as expected). If I do a simple regression on this data I get a function that very closely predicts the records. However, since this graph is linear and positive (not stationary) I can’t just do a regular regression, I have to make it stationary first.
OK so to get to stationarity, I take the first differences and I get a process that looks much more like white noise. I graph the correlogram to confirm stationarity using for k =1, 2, …, and my values fluctuate between -5 and +7.5(no where near 1/N^.5 ). The examples of these graphs of autocorrelation functions are in a much smaller range. I’m using the first difference values for my Yt and to solve for my Y bar. Whats going so wrong?
Putting my confusion of confirmation of stationarity aside, I move on to attempt to graph in excel an ARIMA(1,1,0) and ARIMA(2,1,0) model. Now I have no idea if I’m supposed to be graphing t against Yt, T against the first difference, or what. My graph appears much more useful graphing t against Yt, however I just spent time making this data stationary, so it wouldn’t make sense to use that value. Then to graph the ARIMA(2,1,0) model, I try to do a regression with my data using t as my input, and 1st and 2nd differences as my output, however the regression analysis tool says that I can only have 1 column of input data.
Please help