random number from a data set with a specified error

Hi All, I am trying to generate a set of random numbers from a given distribution. The randn option gives a set of values with a definite mean and standard deviation. However, my intention is to generate several sets of random numbers from a known set of data within a specified error. Please suggest. Atanu

Answer 1

[Geoff]

This question is open to interpretation.

What I think you have is a dataset X with known standard deviations S (a vector containing one deviation for each value in X), and you want to generate a whole bunch of pseudo-X datasets that are randomised the known distributions of X.

If that's the case, you can generate a single set as follows:

N = length(X);
pX = X + S .* randn(1,N);

Repeat as many times as you like.

The error will be normally distributed around each point in X by the known standard deviation at that point.

If you have an error rather than standard deviation, and want random noise within that range, you might do it with a linear distribution instead of normal (let's assume E is an absolute error value, so we're varying within X +/- E):

pX = X + (rand(1,N)*2-1) .* E;

-g-

[Atanu]

Hi,

Here is the matlab file I am using to generate 100 different profiles from the distribution that I have as zndep. I have used an error of 1. Now I want to use the error value of 2.7. How could I do that? Both hist-zndpl.dat and hist-zndpl-cu.dat has one column of data each.

Thanks

Atanu

% errorcheck.m in ZN-dep and ZN-dep-cuShift

clear, close all,

zndep = load('hist-zndpl.dat');

zndepcushift = load('hist-zndpl-cu.dat');

zndep = sort(zndep);

zndepcushift = sort(zndepcushift);

f = (1:length(zndep))./length(zndep) ;

plot(zndep,f,'.-g'), hold on

plot(zndepcushift,f,'.-r')

% now take gaussian error bars around zndep with appropriate %width sig=1kJ/mol

for i=1:100

r = randn(length(zndep),1); % rand #s with sig=1

zni = zndep + r;

plot(sort(zni),f,'-g')

end

[Geoff]

r = 2.7 * randn(length(zndep),1);

[Atanu]

Thanks a lot.