I am currently in time series with my MacBook Pro paying attention to the current model we are discussing-the Box-Jenkins ARMAX Model (aka the Transfer Function Model). Basically you can have two different kinds of models. One, a systematic model which has your basic observed inputs at time n and observed ouputs at time n. The other model is called a disturbance model which has a disturbance at time n and a white noise error term at time n. Then all of a sudden, I wondered how I would model graduate students in the doctoral program. [Looking back on this: I realize potentially how ridiculous this sounds in my blog. But I think this is a legitimate question and potentially interesting question.]
But then I thought back to the stochastic different equation model because one assumption of the model is that you don't receive any feedback, except for recursive forms. The systematic model has a feedback term but only has 0, 1, or 2 higher order terms which maybe could be thought of in terms of years? But then if you combine 2 years for each order and take longer than 6, then you're screwed. Would there be enough of a stationary process to occur in the disturbance model during your graduate school life that may actually be so random and chaotic? But then I thought about the backshifts that occur in the model, however the lags might be appropriate. Not sure. Then I thought about all the different pathways that graduate students take for their doctorate and then I thought about this great quote I heard in my other statistics class, linear models:
All models are wrong, but some are useful.
This ended my thinking about creating a time series model for graduate students.

My recent trip to Washington included a day of Chris and I running around trying to see everything DC has to offer.