我正在尝试从八度运行M文件,但出现此错误:
Octave evaluation error: 'ncx2cdf' undefined
Apparently the non-central chi-square distribution is simply defined as 1 minus the Marcum Q function . The signal package at octave-forge provides an implementation for this function (seemingly compatible with matlab).
Therefore you could presumably write your own ncx2cdf
function simply as follows:
function Out = myncx2cdf (X, V, Delta)
Out = 1 - marcumq (sqrt (Delta), sqrt (X), V/2);
end
Confirmed in matlab:
>> X = randi(100, [1,20]); V = 4; Delta = 10;
>> ncx2cdf(X, V, Delta)
ans =
1.0000 0.9410 0.9999 1.0000 1.0000 1.0000 1.0000 0.5549 0.6093 0.9410 1.0000 0.9410 1.0000 0.9279 1.0000 0.9920 0.8183 0.9410 1.0000 0.9997
>> 1 - marcumq(sqrt(Delta), sqrt(X), V/2)
ans =
1.0000 0.9410 0.9999 1.0000 1.0000 1.0000 1.0000 0.5549 0.6093 0.9410 1.0000 0.9410 1.0000 0.9279 1.0000 0.9920 0.8183 0.9410 1.0000 0.9997
Octave session for the same X, V, and Delta:
octave:34> pkg load signal
octave:35> 1 - marcumq(sqrt(Delta), sqrt(X), V/2)
ans =
1.00000 0.94105 0.99988 1.00000 1.00000 1.00000 0.99996 0.55492 0.60929 0.94105 1.00000 0.94105 1.00000 0.92793 1.00000 0.99203 0.81831 0.94105 1.00000 0.99972
Note that the degrees of freedom parameter V is restricted to even values with this implementation; if you'd like to use odd degrees of freedom too, eg 5, this could be interpolated from the result for V=4 and V=6 (this seems to work well in practice).
here is a easy implementation of non-central chi square distribution:
function f = ncx2pdf(x, n, lambda, term = 32)
f = exp(-lambda/2) * arrayfun(@(x) sum_expression([0:term],x,n,lambda), x);
function t = sum_expression(j,v,n,l)
# j is vector, v is scalar.
numerator = (l/2).^j .* v.^(n/2+j-1) * exp(-v/2);
denominator = factorial(j) .* 2.^(n/2+j) .* gamma(n/2+j);
t = sum(numerator ./ denominator);
end
end
here is the function file , put it in your octave path.
The technical post webpages of this site follow the CC BY-SA 4.0 protocol. If you need to reprint, please indicate the site URL or the original address.Any question please contact:yoyou2525@163.com.