---

---

4. Preparing data

A special input file specifying the type and the distribution of the state vector (see Section 2) should be prepared before running the code.

The k subsequent groups of i.i.d. entries of Y are specified in this file one after another, and fragment describing pair ID[i],v_i specifying i-th group of i.i.d. entries consists of m_i+1 subsequent lines of the data file, m_i being the number of parameters of the pdf f_i.

The syntax of the data file is clear from the following example:

* This is a comment line (asterisk in the first position).
* Such a line will be ignored
BT 4
* The above line says that the first four (the number after 'BT')
* components in the state vector are identically Beta-distributed
* on (0,1)
0.5 1 1.5
* the first parameter of Beta is allowed to vary from 0.5 to 1.5,
* the actual value being 1.
0.5 1 1.5
* the second parameter of Beta is allowed to vary from 0.5 to 1.5,
* the actual value being 1
* Note that the description of the range and the actual value of
* a parameter always occupies exactly one line
DS 30 8
* The above line says that the next 30 components in the state
* vector are identically distributed on the 8-point set 
* {0,1,2,3,4,5,6,7}
0.01 0.125 0.99
* the probability to get 1 is allowed to vary from 0.01 to 0.99,
* the actual value being 0.125
0.01 0.125 0.99
* similarly probability to get 2
0.01 0.125 0.99
0.01 0.125 0.99
0.01 0.125 0.99
0.01 0.125 0.99
0.01 0.125 0.99
0.01 0.125 0.99
* similarly probabilities to get 3 - 8
BN 1
* The above line says that the next component in the state vector is
* Bernoulli distributed
0.01 0.5 0.99
* the probability to get 1 is allowed to vary from 0.01 to 0.99, the
* actual value being 0.5
GS 2
* The above line says that the next two components of the state vector
* are Gaussian random numbers
0 0 0
* the mean of these Gaussian random numbers is fixed to be 0
0.01 1 100
* the standard deviation of these Gaussian random numbers is allowed
* to vary from 0.01 to 100, the actual value being 1
RL 3
* The above line says that the next 3 components in the state vector
* are Rayleigh distributed
0.01 1 100
* the parameter of the Rayleigh distribution is allowed to vary from 
* 0.01 to 100, the actual value being 1
$$
* The above line marks the end of the data file

Range of a parameter.

In the above description actual value of every distribution parameter is accompanied by lower and upper bounds on the parameter, so that the description of a parameter is

 lower_bound actual_value upper_bound,

where

lower_bound < actual_value < upper_bound.

The segment [lower_bound,upper_bound] is called the range of the corresponding distribution parameter; the code looks for the value of the related reference parameter exactly in this range, and one can use this fact in order to specify a desired "direction of search" (e.g., in the reliability analysis of a network with Bernoulli distributions of arc failures, one can assume a priori that there is no necessity to use reference parameters corresponding to "more reliable" arcs than the actul ones; to forward this instruction to ASFE, it suffices to specify lower bounds on failure probabilities to be equal to the actual values of these probabilities.

For part of distribution parameters, the lower bounds can coinside with the upper ones (and, consequently, with the actual values of the parameters). The corresponding parameters are called fixed, and the corresponding entries in the reference parameter vector will be the same as in the actual parameter vector.

Remark 4.1

By serious reasons, for every Normal distribution (type GS) at least one of the underlying parameters - or - must be specified as fixed.

---

---