1
$\begingroup$

I read the following paragraph which claims to be the Taylor expansion of standard normal CDF for positive $x$.

$1 - \frac{1}{\sqrt{2\pi}}e^{-\frac{x^2}{2}} (a_1k + a_2k^2 + a_3k^3 + a_4k^4 + a_5k^5)$,

where $k:=\frac{1}{1+0.2316419x}, a_1=0.319381530, a_2=-0.356563782, a_3=1.781477937, a_4=-1.821255978, a_5=1.330274429.$

I do not get how this is derived even though I am numerically convinced of its truth. Could anyone explain this to me, please?

$\endgroup$
2
$\begingroup$

This is no Taylor expansion. It is a rational approximation listed as formula 26.2.17 in Abramowitz / Stegun (see http://people.math.sfu.ca/~cbm/aands/page_932.htm) and attributed to Hastings (1955), Approximations for Digital Computers.

$\endgroup$
  • $\begingroup$ Thanks. Where to find how this is derived, please? $\endgroup$ – LaTeXFan Nov 9 '16 at 10:17
  • $\begingroup$ @LaTeXFan. Probably curve fit after having chosen the form of the approximation. $\endgroup$ – Claude Leibovici Nov 9 '16 at 10:21
  • $\begingroup$ May be you have access to the Hastings reference in a public library, I do not know a derivation. $\endgroup$ – gammatester Nov 9 '16 at 10:22

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.