Mathematica 9 is now available
Services & Resources / Wolfram Forums
-----
 /
MathGroup Archive
1996
*January
*February
*March
*April
*May
*June
*July
*August
*September
*October
*November
*December
*Archive Index
*Ask about this page
*Print this page
*Give us feedback
*Sign up for the Wolfram Insider

MathGroup Archive 1996

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

Search the Archive

Accuracy.

  • To: mathgroup at smc.vnet.net
  • Subject: [mg5038] Accuracy.
  • From: Hyunho Shin <shinhh at plaza.snu.ac.kr>
  • Date: Sat, 19 Oct 1996 16:40:38 -0400
  • Organization: School of Economics, SNU.
  • Sender: owner-wri-mathgroup at wolfram.com

Hi,

In Mathematica, I run the next:

----------------------------------------------------------------------
<<Statistics`Master`

n = 500;

x2dist = NormalDistribution[1,1];
x3dist = NormalDistribution[-1,4];
udist = NormalDistribution[0,1];

x1 = Table[1, {n}];
x2 = Table[Random[x2dist], {n}];
x3 = Table[Random[x3dist], {n}];
u = Table[Random[udist], {n}];
x = Transpose[{x1, x2, x3}];
ystar = x.{1,2,3} + u;
y = Table[If[Part[ystar, i] > 0, 1, 0], {i, n}];

tol = 0.00000001;
b0 = {1,1,1};

Do[
   Print[b0];
   f = PDF[NormalDistribution[0,1], x.b0];
   F = CDF[NormalDistribution[0,1], x.b0];
   first = Inverse[Transpose[(f^2/(F(1-F)))*x].x];
   second = Transpose[(f/(F(1-F)))*x].(y-F+f*x.b0);
   b1 = first*second;
   If[Max[Abs[b0-b1]/bo] < tol, Break[]];
   b0 = b1,
{100}];

"In PROBIT Model,"
"Estimated Parameter: bprobit = " b0 ", and"

fprobit = PDF[NormalDistribution[0,1], Map[Mean, Transpose[x]].b1];

"(d F)/(d xj) = " fprobit * b0 
--------------------------------------------------------------------

But I got the message:
                                     1
    Power::infy: Infinite expression -- encountered.
                                     0.
because, in Mathematica, some elements of F are taken to 1.
But, F is a list of values of CDF of Standard Normal Distribution,
so any element of F cannot be 1.

I think Mathematica round off all elements of F,
some values, eg 0.99999999999999999999999999999999999999991, are
taken to 1's.

I want the method in which I get sufficiently accruate values,
in this context.

Thanks in advance,

Hyunho Shin (shinhh at plaza.snu.ac.kr)


  • Prev by Date: Re: boundary value problems
  • Next by Date: Why no solution to 3 eqns. in 3 unk. ?
  • Previous by thread: Comment: Fit in Excel Mathlink & some
  • Next by thread: Why no solution to 3 eqns. in 3 unk. ?