[R] Trouble understanding the behaviour of stableFit(fBasics)

[Ted Byers]

Can anyone explain such different output:

> stableFit(s,alpha = 1.75, beta = 0, gamma = 1, delta = 0, 
+     type = c("q", "mle"), doplot = TRUE, trace = FALSE, title = NULL, 
+     description = NULL)

Title:
 Stable Parameter Estimation 

Call:
 .qStableFit(x = x, doplot = doplot, title = title, description =
description)

Model:
 Student-t Distribution

Estimated Parameter(s):
     alpha       beta      gamma      delta 
 1.5340000  0.2750000  0.3211991 -0.9922306 

Description:
 Tue Sep 23 22:18:44 2008 by user: Ted 

> refdata18 = read.csv("C:\\MerchantData\\RiskModel\\Capture.Week.18.csv",
> na.strings="")
> stableFit(refdata18[,1],alpha = 1.75, beta = 0, gamma = 1, delta = 0, 
+     type = c("q", "mle"), doplot = TRUE, trace = FALSE, title = NULL, 
+     description = NULL)

Title:
 Stable Parameter Estimation 

Call:
 .qStableFit(x = x, doplot = doplot, title = title, description =
description)

Model:
 Student-t Distribution

Estimated Parameter(s):
alpha  beta gamma delta 
   NA    NA    NA    NA 

Description:
 Tue Sep 23 22:20:23 2008 by user: Ted 

> 


I am just playing with it right now, trying to understand how to call it, so
first I passed the s vector from the example.  I don't care about the result
except to know that stableFit accepted the input and obtained an estimate
for the parameters.

The I tried my data (a vector in integers, with a distribution that looks
similar to  poisson, but exponential and geometric give better fits).

What I find puzzling is that I get no error messages complaining about one
property or another of my data, to explain why there are no parameter
estimates.  The data I WILL be applying this to comes from the financial
markets, and will be reals or floating point numbers that in some cases wil
be best modelled by a normal distribution while in most cases, the
distribution will be closer to cauchy.  (but DistributionFits(fBasics) makes
no explicit mention of cauchy,  but IIRC cauchy is a special case of a
stable distribution one of a family - are these the L-stable distributions
Mandelbrot discussed, or something else - correct me if my memory has failed
me sooner than anticipated ;-)  An URL for a website discussing these in
some detail would be handy as my stats texts, dated as they are and focussed
more on applied biometrics, don't talk about these.

What do I look at if this function just gives me a bunch of 'NA's instead of
parameter estimates?

And, givent he structure of the documentation, it is not clear if I can get
an estimate of skewness for all the distributions or for all except t and
normal distributions if I am using DistributionFits.

Thanks

Ted
-- 
View this message in context: http://www.nabble.com/Trouble-understanding-the-behaviour-of-stableFit%28fBasics%29-tp19640972p19640972.html
Sent from the R help mailing list archive at Nabble.com.