MathGroup Archive 2003

[Date Index] [Thread Index] [Author Index]

Search the Archive

Re: WeibullDistribution

  • To: mathgroup at smc.vnet.net
  • Subject: [mg42460] Re: WeibullDistribution
  • From: Bill Rowe <listuser at earthlink.net>
  • Date: Wed, 9 Jul 2003 08:24:32 -0400 (EDT)
  • Sender: owner-wri-mathgroup at wolfram.com

On 7/8/03 at 4:37 AM, ce.choa.phen.kee at philips.com wrote:

> I have a set of data, but how can I find out the A and B in 
> WeibullDistribution[ A , B ] ?

There are a number of ways to do this. One simple way (perhaps not the best) would be to compute A and B from two given quantiles, i.e., noting for a Weibull distribution

Log[-Log[1-p] == A Log[t] - A Log[B]

where t is the p_th quantile.

Using two quantiles in this manner gives essentially a point estimate of A and B,

A more global estimate can be obtained using Fit to do a regression analysis of Log[data] vs Log[-Log[1-p]] where each data point is paired with its corresponding quantile.

There are a number of choices that can be used for p. Probably the most commonly used is (j-1)/n. This can be computed from a vector of data points as

MapIndexed[(First@#2-1)/Length[data]&,Sort@data]

Other methods for computing A are based on maximum liklihood methods or moments of Log[data]. Refer to a good statistics text for details.


  • Prev by Date: Re: WeibullDistribution
  • Next by Date: NDSolve: Precision and Stability
  • Previous by thread: Re: WeibullDistribution
  • Next by thread: Re: WeibullDistribution