Can anyone help me with the construction of confidence and prediction
intervals for time series forecasts?
I am familiar with CIs and PIs for estimated values of y ( y = B0 + B1x)
with ordinary least squares regression. In this case, I have a naive time
series model that has been transformed to a linear relationship and
estimated with OLS.
model: Y-stock = N0 - N1*exp{-kt}
==> Y-increment = k*N1*exp{-kt}
estimate ln{Y-increment} = ln{k*N1} - kt
Intervals in OLS are a function of the variance of the estimate, which
increases as one's prediction moves further away from the mean of the x
values. Is this same process applicable with this model? How can I
construct such an interval to splice on the end of the data series to
provide some idea of forecase error?
Many thanks.
David Petersen
