3/6/2006

The small glitch results indicate that Gauss Laguerre is the most accurate routine for z<-2.5.

The codes

Figure 1 Relative error as a function of z for AiGauss(z)

The error estimates below -2.6 are from 19 – 20 point Gauss Laguerre quadratures. Those above are from 6 - 7 point Gauss Legendre quadratures.

The code is genint.wpj

02/26/2006 08:02p 772 TBRACK.FOR – used for asymptotic expansion test

02/26/2006 08:58p 557 AIGZM.FOR -- used for asymptotic exapansion

02/22/2006 02:49p 181 TAIGSM.FOR – used for asymptotic expansion

02/28/2006 05:03a 1,119 TGLEG67.FOR – old method not quite as accurate

03/06/2006 08:21a 980 genint.for – main code

02/26/2006 08:52p 978 GLEG67.FOR -- used for Gauss Legendre integration

03/05/2006 08:19p 2,277 GLAGU.FOR – copied from aiglz

03/06/2006 04:41a 476 BRGLAGU.FOR – copied from aiglz