Determine the number of required Monte Carlo iterations

[ryan88]

I am performing a Monte Carlo simulation of a weapon system to determine the effect of variations in weapon characterisitics on the accuracy of the weapon. For example, I might vary the muzzle velocity of the weapon and measure the angles between the aim point and the impact point (i.e. miss angles).

I need to report the standard deviation of the miss angles in milli radians to two decimal places, so I would like to determine the minimum number of Monte Carlo iterations required to obtain a 95% confidence level that the standard deviation of the miss angles are accurate to 0.01 mili radians.

My statistics are a bit rusty, so does anyone know of the best way to go about this?

 

[mfb]

The uncertainty on the standard deviation goes down with the square root of the number of data points (as usual), so it depends on the size of your standard deviation and some numerical prefactor you can look up.