Differentiating a cumulative distribution function

Question

Carter 2012-9-17

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/48385-differentiating-a-cumulative-distribution-function

I am trying to numerically evaluate the integral of the normal density function of x with respect to the cumulative distribution of (x+y-m).

I am currently attempting to solve this by trying to integrate the derivative of the cdf times the pdf with respect to x.

Here is my code so far:

k=100;
m=20;
y=0;
syms x;
f=normpdf(x);
g=diff(normcdf(y-x+m);
D=simple(int(f*g,0,(k-m)))
Q=double(D);
E=normcdf(y+2*m-k)-D

Any suggestions? The error message is Error using NaN Trailing string input must be 'single' or 'double'.

Error in normcdf (line 60)
p = NaN(size(z),class(z));

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Tom Lane 2012-9-17

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/48385-differentiating-a-cumulative-distribution-function#answer_59123

在 MATLAB Online 中打开

The Statistics Toolbox functions normpdf and normcdf don't accept symbolic inputs. You might be able to use functions like erf, which do operate on symbolic variables, to accomplish what you want to do. Here's a little example where I compute normcdf on a numeric vector, then I use Symbolic Toolbox functions to get the same result:

>> x = [.5 .7 .9];
>> normcdf(x)
ans =
    0.6915    0.7580    0.8159
>> syms s;
>> p = erfc(-s/sqrt(2))/2;
>> double(subs(p,'s',x))
ans =
    0.6915    0.7580    0.8159

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Carter 2012-9-27

Thanks, as I've stated below, this works, however it is through trial and error not the way I am wanting to go about this.

请先登录，再进行评论。

Answer 2

Carter 2012-9-27

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/48385-differentiating-a-cumulative-distribution-function#answer_60108

在 MATLAB Online 中打开

Thanks! You've really helped me out there. However, I've realized I need to go about this same thing differently. Learning from what you've shown,

syms x mu sigma
f=1/(sigma.*sqrt(2.*pi)).*int(exp(-(x-mu)^2)./(2.*(sigma^2)),-inf,x);
subs(f,{x mu sigma},{3.75 3.75 1}) % Should be equal to below
normcdf(3.75,3.75,1) % normcdf

I however do not get the same answer (should be .5 in this example). What did I do wrong?