home *** CD-ROM | disk | FTP | other *** search
- " z erf(z) abs(error)
- "
- 0.0 0.0
- 2.579182479E-11 2.579182479E-11 {transition point between 0 and
- 2.579282480E-11 2.579282480E-11 non-zero for CP/M Turbo Pascal}
- 1E-10 1E-10 {for extremely small values,
- 1E-6 1E-6 erf(x) = x to within accuracy}
- "
- .01 .0112834156
- .02 .0225645747
- .03 .0338412223
- .05 .0563719778
- .06 .0676215944
- .08 .0900781258
- .09 .1012805939
- "
- " The values computed for erf(z), with z small, have the
- " indicated variance from table values of The Handbook
- " of Mathematical Functions [HMF: Table 7.1]. Results
- " to within 1E-5 are usually quite acceptable.
- / z erf(z) abs(error)
- "
- .18 .2009358390
- .37 .3992059840
- "
- .46 .4846553900
- .47 .4937450509
- .48 .5027496707
- .49 .5116682612
- .50 .5204998778
- "
- .60 .6038560908
- .81 .7480032806
- "
- " Midrange values, near .5, reveal the best that can be
- " obtained when either erf(z) or erfc(z) = 1-erf(c) is
- " used. Test values are from [HMF: Table 7.1].
- / z erf(z) abs(error)
- "
- 1.83 .9903468051
- 2.506628275 .9996072495
- 3.069980124 .9999858556
- 3.544907702 .9999994648
- 3.963327298 .9999999792
- 4.341607527 .9999999992
- "
- 4.6894721 1.0
- 5.0 1.0
- 10.0 1.0
- "
- " Near the tail, as erf(z) approaches 1.0, [HMF: Table 7.3]
- " values of erfc(z) = 1 - erf(z) are used to see how well
- " accuracy is maintained despite adjustments by 1.0 in erf.
- "
- "ERFT1> End of test.