Thanks for your advice. Another question.
One of my friends told me one way to simulate normal variables. it is..
Y1,Y2: random samples from the uniform distribution over 0 < y < 1
X1,X2 can be got by
X1 = SQRT(-2 * LN(Y1) * COS(2 * 3.141592 * Y2)
X2 = SQRT(-2 * LN(Y2) * COS(2 * 3.141592 * Y1)
He found it from "Introduction to Mathematical Statistics 4th ed." by Robert
V. Hogg & Allen T. Craig (The university of Iowa) and he said that that way
was suggested by Box and Muller.
In this way, NORMSINV() can be correctly implemented? I tried to plot two
histograms with excel. One is by this way and the other is by NORMSINV()
with 10000 uniform random numbers from 0 to 1. Two histograms have some
resemblance but not same.
I need any comments. Is it good enough to use? Or should i find another way?
Thanks in advance..
-hk from Korea
%%PS:i cross-posted to groups sci.stat.math,sci.stat.edu. and I am very much
obliged for your mails and postings from Ivan and Alan and Jay.
