0

I was wondering if there is any benefit to using the inverse error function as opposed to the inverse normal in simulating normal random variates. Increased accuracy for probabilities way in the tails would be nice.

I was not able to find any information about this online so thanks!

debo
  • 156
  • 6
  • If the inverse normal is well implemented you should be able to use it with a small fiddle for dealing the upper tail that should give you essentially equivalent accuracy. – Glen_b Jul 25 '17 at 01:19
  • Since the two functions are essentially the same, what exactly are you trying to ask? – whuber Jul 25 '17 at 01:52
  • Matlab gives some tips without justification that using the error function provides increased accuracy in some cases. I was wondering if there is something documented about when it should be used or if it is always preferable. – debo Jul 25 '17 at 03:36
  • Please state the exact tips from MATLAB. Provide links as appropriate. – Mark L. Stone Jul 25 '17 at 11:22
  • Here is the link, the tip involves some typeset math, https://www.mathworks.com/help/matlab/ref/erf.html – debo Jul 25 '17 at 20:28

0 Answers0