ABSTRACT
Computer generation of random variables, no matter how complex it can be, is based on the generation of very simple distributions and further applications of suitable transformations. In practical implementation of algorithms, a random number generator is represented by a deterministic sequence of real numbers between 0 and 1 in the double-precision format. This is very convenient in the first stage of writing a simulation program, where generating truly random results makes finding coding errors difficult. In practice, a generator is cyclic and after L calls it again returns the initial value, etc. Obviously it is important to ensure that the period L of the generator that is used is large enough compared with the number of calls to be made: in practice, most of the currently available generators satisfy this constraint.
