How can I make an Gaussian random array with a constraint?

I'm trying to create an array with size of nPopulation=100. I want all the values to be driven from the standard normal distribution and positive. but sum of the values should be landaT=25000. I've written the code below

flag=0;
while (flag==0)
    landa=(randn(1,nPopulation))*(landaT./nPopulation);
    landa(1,nPopulation)=landaT-sum(landa)-landa(1,nPopulation);

      if sum (landa)==landaT
          flag=1;
      end
  end

But it doesn't seem to work. If you have any better Ideas, it would be great to hear from.

Answer 1

[Orion]

Your criteria is too strict.

it's allways risky to compare 2 doubles, because of the numerical precision.

yo should do something like :

flag=0;
landaT = 25000;
nPopulation=100;
NbIter = 0;
while (flag==0)
    NbIter = NbIter + 1;
    landa=(randn(1,nPopulation))*(landaT./nPopulation);
    landa(1,nPopulation)=landaT-sum(landa)-landa(1,nPopulation);
    if abs(sum(landa)-landaT)<0.01
        flag=1;
    end
end
disp('Number of Iteration to convergence');NbIter

Here, you compare the difference of the 2 parameters you want, in absolute value, to a epsilon you defined (0.01)

[shima rashidi]

Thanks for the answer, its helpful. but my another problem is making the values positive. how can I make them positive and be sure that they still are Gaussian rand numbers.

[Torsten]

You can't because the standard normal distribution takes positive and negative values.

Maybe you mean

http://en.wikipedia.org/wiki/Half-normal_distribution

or

http://en.wikipedia.org/wiki/Truncated_normal_distribution

Best wishes

Torsten.