[R] Value of Hurst exponent (R/S) method > 1

Barun Saha barun.saha04 at gmail.com
Wed May 9 13:58:29 CEST 2012


Hello All,

I'm coming across multiple data sets for which the R/S estimate of H
is greater than 1. Could someone please explain this to me?

On Mon, May 7, 2012 at 4:13 PM, Barun Saha <barun.saha04 at gmail.com> wrote:
>
> Hello,
>
> I'm using fArma package to estimate the value of Hurst exponent using R/S
> method. However, for a certain set of data I get H ~ 1.8. How do I interpret
> this?
>
> Following are the output that I get for this set:
>
> > mean(data[,2])
> [1] 400.5433
> > sd(data[,2])
> [1] 1139.786
> >
> > rsFit(data[,2], levels = 64)
>
> Title:
>  Hurst Exponent from R/S Method
>
> Call:
>  rsFit(x = data[, 2], levels = 64)
>
> Method:
>  R/S Method
>
> Hurst Exponent:
>          H     beta
>   1.826240 1.826240
>
> Hurst Exponent Diagnostic:
>    Estimate  Std.Err   t-value  Pr(>|t|)
> X 1.826240 3.352896 0.5446755 0.5919988
>
> Parameter Settings:
>        n   levels  minnpts cut.off1 cut.off2
>    11895       64        3        5      316
>
> Description:
>  Mon May  7 16:16:59 2012 by user:
>
> Warning messages:
> 1: In rsFit(data[, 2], levels = 64) :
>   Integer overflow in 'cumsum'; use 'cumsum(as.numeric(.))'
> 2: In (1:m) * Y[m] : NAs produced by integer overflow
> 3: In (1:m) * Y[m] : NAs produced by integer overflow
> 4: In (1:m) * Y[m] : NAs produced by integer overflow
> 5: In (1:m) * Y[m] : NAs produced by integer overflow
> 6: In lsfit(log10(M), log10(RS), wt) : 37 missing values deleted
> >
>
>
> --
> Thanks,
> Barun Saha
> JPA
> IIT, Kharagpur
>
> http://pothi.com/pothi/book/barun-saha-swapner-kheya
> http://delay-tolerant-networks.blogspot.com/p/one-tutorial.html
>



--
Thanks,
Barun Saha
JPA
IIT, Kharagpur

http://pothi.com/pothi/book/barun-saha-swapner-kheya
http://delay-tolerant-networks.blogspot.com/p/one-tutorial.html



More information about the R-help mailing list