Hi,
I believe what you want to test for is first order stationarity. The
conditions for this kind of stationarity are that:
1. E[x_t]=m for all t,
2. Variance of x_t does not diverge
3. COV (x_t,x_(t-k)) = gk, for all t
- You could be very lazy and look at the graph. See if the line does not
spend "too much time" above or below the mean value. Then split your sample
in two and see if the two means are the same... I do not recommand this
method, but if you're in a hurry, you usualy get it right (don't tell any of
my professors)
- You could be a little less lazy and take your code to R, d/L the Rmetrics
library from the CRAN site and enjoy the work of others. But if you're like
me and you prefer Splus to R, and well... didn't you pay for Splus at
first?, Well you might want either buy finmetrics or do the next
proposition:
- Calculate your very own ADF test with the lm() function and a lot of
frustration. You can find plenty of doc on the dickey-fuller test over the
internet. However, there are a few things you want to look for.
1. it is important to consider the right number of lags. Not enough means
you OLS is biased, too many and you lose valuable degrees of freedom. I
recommand using Campbell perron procedure.
2. The asymptotic laws become degenerate when your time series are almost
unit root. The t-statistic do not have the same confidence interval as for
normal law. At a 5% IC, those are the t-stat values you should consider.
tstat (no constant, no trend) : -1.94
tstat (constant, no trend) : -2.86
tstat (constant, trend) : -3,41
For more information on the Campbell Perron proc. or the Adf test, I
recommend you look for "TimeSeries Analysis" by Hamilton, but boy is this
hard work! Otherwise, find a professor that teaches Macroeconometrics and
check out his script. I would recommand Alain Guay, but his script is in
french...
I hope this could help.
Lambert.Winnie wrote:
>
> All,
>
>
>
> Is there a function in S-PLUS that can be used to test whether a time
> series is stationary or not? Thanks.
>
>
>
> *****************************************************************
>
> Winifred C. Lambert Senior Scientist/Meteorologist
>
> ENSCO, Inc.
>
> Aerospace Sciences and Engineering Division
>
> 1980 N. Atlantic Ave., Suite 230
>
> Cocoa Beach, FL 32931
>
> VOICE: 321.853.8130 FAX: 321.853.8415
>
> EMAIL: lambert.winnie@ensco.com <mailto:lambert.winnie@ensco.com>
>
>
>
> AMU Quarterly Reports are available online:
>
> http://science.ksc.nasa.gov/amu <http://science.ksc.nasa.gov/amu>
>
> *****************************************************************
>
> ......................................................................
> The information contained in this email message is intended only for the
> use of the individuals to whom it is
> addressed and may contain information that is privileged and sensitive. If
> the reader of this message is not
> the intended recipient, you are hereby notified that any dissemination,
> distribution or copying of this
> communication is strictly prohibited. If you have received this
> communication in error, please notify the
> sender immediately by email at the above referenced address. Thank you.
>
--
View this message in context:
http://www.nabble.com/stationary%2C-or-not-stationary%3A--time-series-dilemma-tp15293311p15297181.html
Sent from the S-News mailing list archive at Nabble.com.
|