The RATS Software Forum

Hi,
I have a little problem for computing modified R/S Statistic.
To estimate the Hurst exponent by the modified R/S Statistic (proposed by Lo), I am trying to use RSStatistic synthax. But, when I write, for exemple, “@ RSSTatistic (lags = m) returnseries start end”, I get no output unlike the synthax “@hurst returnseries” with which I get directly the Hurst and d exponents. How can I compute the Hurst exponent by the modified R / S? Where I make an error?
Thank you for your help.

The RSStatistic procedure doesn't produce any output since you usually do quite a few of them (varying the range or the window width) and produce some type of sensitivity table. The value of the R/S statistic is put into the %CDSTAT variable. The Willinger, Taqqu and Teverovsky replication file has an example using it.

TomDoan wrote:The RSStatistic procedure doesn't produce any output since you usually do quite a few of them (varying the range or the window width) and produce some type of sensitivity table. The value of the R/S statistic is put into the %CDSTAT variable. The Willinger, Taqqu and Teverovsky replication file has an example using it.

Thank you for your reply.
Is their 1999 paper that you advise me?
In fact, when I write "@ RSStatistic (lags = k) Rt 1 5300" (I compute in advance the lags k by the formula of Andrews (1991)), and I write "tables", I get the statistics of my variable Rt and then variables "DX" and "SUMDX. Which is the value of the modified R / S? I do not know exactly how I have to write on the syntax on RATS for the modified R / S statistic. Can you please help me?
Thanks.

%CDSTAT is a scalar, so you use DISPLAY:

TomDoan wrote:%CDSTAT is a scalar, so you use DISPLAY:

Code: Select all

@RSStatistic(lags=k) Rt 1 5300
disp "R/S Statistic" %cdstat

Thank you for your help.

Hi
i am using the source code HURST for estimating hurst exponent. i want to calculate V-statistics. please advise if it is possible using the same source file.

ritesh wrote:Hi
i am using the source code HURST for estimating hurst exponent. i want to calculate V-statistics. please advise if it is possible using the same source file.

Just to understand what you're asking, instead or (or in addition to) the log obs vs log R/S graph, you want the same thing but with Lo's modified statistic? If you just want to calculate that for the full data set, you can use the RSStatistic procedure. What the Hurst procedure computes are the R/S statistics for many different subsample sizes, and for each possible subsample of each size.