Suppose you fit a time series model by differencing the outcome variable yt and the input variable xt and includes lags, such as ∇yt = β1∇yt−1 + β2∇xt + β3∇xt−1 + t. where ∇ is the first differences operator. For this model,
(a) What assumptions are you making about the dynamic relationship of yt and xt in fitting such a model?
(b) What happens to the properties of your dynamic model estimates if the errors are still serially correlated versus when they are white noise?
(c) What restrictions are you imposing on the dynamic relationships? Are these valid? How coul