I am working on a problem where I want to backcast a time series, i.e.,
forecast into the past to a time before the measurements of the times
series were available.
None of the software packages I have access to (including Gretl) appear
to offer an explicit backcasting option. I have read that a stationary
time series is time-invariant, i.e., it has the same properties forwards
and backwards in time. So, when I go to backcast the data is it valid to
analyze the data in Gretl in reverse order and just do a regular forecast?
Tom