Mathematica 9 is now available
Services & Resources / Wolfram Forums / MathGroup Archive
-----

MathGroup Archive 2012

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

Search the Archive

Re: "Bug" in SurvivalDistribution?

  • To: mathgroup at smc.vnet.net
  • Subject: [mg127048] Re: "Bug" in SurvivalDistribution?
  • From: Darren Glosemeyer <darreng at wolfram.com>
  • Date: Wed, 27 Jun 2012 04:07:42 -0400 (EDT)
  • Delivered-to: l-mathgroup@mail-archive0.wolfram.com
  • References: <201206260848.EAA06907@smc.vnet.net>

On 6/26/2012 3:48 AM, Mark Coleman wrote:
> Greetings,
>
> I've bumped into an unusual condition using the built-in SurvivalDistributi=
> on function in Mathematica v8.0.4. I'm hoping another list member could run this code on their system and let me know if they get the same behavior.
>
> I've been doing some basic exploratory survival analysis and attempted to generate a SurvivalDistribution object for n=1900 event times. On my system (Dell T3400 with 8 Gb Ram running 64-bit Windows 7 Enterprise edition, SP1), the following code works for n<1900:
>
> n = 1899;
> data0 = RandomReal[LogNormalDistribution[1, 5], n];
> survDist = SurvivalDistribution[data0]
> Plot[SurvivalFunction[survDist, z], {z, 1, 100}]
>
> but when n>= 1900, my call to SurvivalDistribution does not evaluate properly.
>
> I've duiplicated this on a much larger Mathematica v8 Linux server. I've also reported it to WRI.
>
> Any idea on what's going on would be most welcomed. My final datasets contain much larger samples, n>100000.
>
> Thanks,
>
> Mark
>

This appears to be fixed in the development version. As a workaround for 
now, you can evaluate the following (it seems the object being created 
is not being flagged as a valid distribution):

Unprotect[DataDistribution];
DataDistribution /:
  HoldPattern[
   DistributionParameterQ[
    DataDistribution["Survival", {_, _}, _, _]]] := True

After that,

n = 1900;
data0 = RandomReal[LogNormalDistribution[1, 5], n];
survDist = SurvivalDistribution[data0];

Plot[SurvivalFunction[survDist, x], {x, 1, 100}]


will work just fine.

Darren Glosemeyer
Wolfram Research




  • Prev by Date: Re: Issue understanding the use of Cashflow
  • Next by Date: Re: Options Question ver 8
  • Previous by thread: Re: "Bug" in SurvivalDistribution?
  • Next by thread: Re: "Bug" in SurvivalDistribution?