[ns] Generating random values having the probability density function

["Kejvan Redjamand (ERV)" <kejvan.redjamand@erv.ericsson.se>]

Hi
I'm a student doing my  Master Thesis. I need to implement a  function which
calculates (actually generates random value of I/S values given its
probability density function, f(x). How should the algorithm look like?
Can any one help me with checking my thoughts below?
 I/S-values which is continuously distributed: f(x). It depends On N, number
of users in our system which can take discrete values. We must know the
typical values for I/S (we call it stochastic variable 'u'). 
 f(x) means actually P(u=x), where x is an actual concrete value from the
values which the stochastic var u can take. 
Then by assuming N form MIN to MAX (e.g.-1000) we can calculate 
	1. probabilities f1, f2,... fmax and store them in an array: P.
	2. I/S values (i.e. concrete u's- u1..umax  corresponding to
probabilities) and store them in array U.
The first question is which u's can we take? (Note that u is takes continuos
values)
Then second question is that we need to use random I/S- values from (U), but
according to their probabilities from P; How can it be done?
The main question is: Is the whole idea reasonable?

regards

// Kejvan
Chalmers University of Technology, Sweden