[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Help verifying statistics distribution functions
From: |
Rik |
Subject: |
Help verifying statistics distribution functions |
Date: |
Sun, 19 Jun 2011 10:58:47 -0700 |
I'm in the process of overhauling the distribution functions in Octave in
order to expand them to accept and return class single arguments. I've
found and fixed a bunch of corner cases, but I need some compatibility
testing from someone with the Matlab Statistics Toolbox.
What do the following code snippets do under Matlab?
A) inv(cdf(x)) == x. This doesn't work for binomial distribution in Octave
binoinv (binocdf (1,4,0.5), 4, 0.5)
B) values outside range return 0 (no probability) or NaN (out of range)?
binopdf (-1, 4, 0.5)
C) Does F distribution accept non-integer degrees of freedom (DOF)
fpdf (1,4.5,2)
D) DOF outside range cause error, or change return values to NaN?
fpdf (1, -1, 2)
E) How are infinite values handled in F distribution
finv (0.5, Inf, 2)
F) How are infinite values handled in gamma distribution
gaminv (0.5, Inf, 1)
G) Possibly incorrect boundary case. Retuns -Inf under Octave but should
be 0?
hygeinv (0, 4, 2, 2)
H) How are non-integers handled by hypergeometric distribution
hygecdf (1, 4.5, 2, 2)
I) How are negative integers handled by hypergeometric distribution
hygecdf (1, -4, 2, 2)
J) How are impossible conditions handled by hypergeometric distribution
hygecdf (3, 4, 2, 2)
K) Does Matlab accept sigma=0 for normal distribution
normpdf (1,1,0)
L) Is empty matrix input acceptable to normal distribution
normcdf ([])
M) Does T distribution accept non-integer DOF
tpdf (1, 1.5)
N) Boundary value for uniform discrete inverse.
unidinv (0, 2:5)
O) How are values not in discrete set treated
unidpdf (-1, 10)
unidpdf (2.5, 10)
unidcdf (3.5, 10)
Thanks in advance for any responses,
Rik
- Help verifying statistics distribution functions,
Rik <=