Normal Quantile with Precision

computes the normal quantile function with high precision for extreme values in the tail

Version 2.0.0.0 (3,24 ko)
269 téléchargements
(1)
27 avr. 2016

Vous suivez désormais cette soumission

computes the quantile function of the standard normal distribution, truncated to the interval [l,u].
Method designed for precision in the tails. Inf values for vectors 'l' and 'u' accepted;
%Example: Suppose you desire the median of Z~N(0,1), truncated to Z>9;
norminvp(0.5,9,Inf) %our method
% in contrast, Matlab's default norminv.m fails
pl=normcdf(9); norminv(0.5*(1-pl)+pl)

Reference:
Botev, Z. I. (2016). "The normal law under linear restrictions:
simulation and estimation via minimax tilting". Journal of the Royal Statistical Society: Series B (Statistical Methodology). doi:10.1111/rssb.12162

For more information, see: https://en.wikipedia.org/wiki/Normal_distribution#Quantile_function

D’autres ont également téléchargé

MITT

806 téléchargements

5,00 / 5 (2)

plotwithmarkers.m

2,7K téléchargements

3,00 / 5 (4)

ballplot2

2,4K téléchargements

3,00 / 5 (2)

Citation pour cette source

Zdravko Botev (2026). Normal Quantile with Precision (https://fr.mathworks.com/matlabcentral/fileexchange/53605-normal-quantile-with-precision), MATLAB Central File Exchange. Extrait(e) le .

Catégories

En savoir plus sur Numbers and Precision dans Help Center et MATLAB Answers

Informations générales

Compatibilité avec les versions de MATLAB

  • Compatible avec toutes les versions

Plateformes compatibles

  • Windows
  • macOS
  • Linux
Ouvrir dans un nouvel onglet
Version Publié le Notes de version
2.0.0.0

- updated the Newton method in "newton(p,l,u)" so that now the exit condition uses
relative error; previous version used absolute error criterion.
1.1.0.0

Vectorized the last update of the function.
If abs(x)>10^6, then there is no need for Newton iteration and the code now bypassed the
iteration;
Faster convergence of Newton's method using
the better initial guess of the root of nonlinear equation:
x=sqrt(l-2*reallog(1+p.*expm1(l/2-u/2)));
Corrected spelling mistake in title
Forgot to suppress output by adding a semicolon.
1.0.0.0

Uploaded better cover picture.
Added more tags.

Added reference to article.
Updated title: "Normal Quantile Function with Precision"