Dear all
I am currently involved in a project which involves constructing
confidence intervals for the mean of a stationary time-series
(serially-correlated data). Does anyone know of any good references
that cover this topic?
One method I have used is to estimate the (auto-) covariance matrix from
the data, and used this to calculate an empirical estimate of the
standard error of the sample mean. But this leads to a problem in that
there are not many pairs of data points available to calculate the
autocovariances at long lags. Has anybody ever looked at this problem
before?
Another method would be to fit say an ARMA model, and use the model to
estimate the standard error of the sample mean, but that would depend on
finding a decent model that fits the data.
Are there any more methods I could use? The next method I will look at
is a bootstrap.
Thanks.
David
NOTICE: Please note that this eMail, and the contents thereof,
is subject to the standard NMMU eMail disclaimer which may be found at:
<http://www.nmmu.ac.za/disclaimer/email.htm>>
|