plan1: generates imposed values to start a simulation plan on a bn

In rebastaba: Handling Bayesian networks

Description

(bn) A matrix of phi(N) values is extracted from (most often marginal) nodes according to the so-called uniform way to serve starting value(s) when simulating from a bn. It is what we called a simulation plan.
The constraints about the nodes to plan are those of the function plan0 which is called (with Nx or Nn) for each node. It is supposed that Nx >= Nn since x is for maXimum and n for miNimum. For the convenience of the explanation, let us denote Vx[i] the set of Nx values associated to the node number ith and Vn[i] the set of Nn values associated to the same node. If Nx==Nn, the two sets are identical (resulting from the same call of plan0) but when Nn < Nx, there is no obligation that Vn be a subset of Vx (see the algorithm in plan0 to get the answer).
Now the plan is constructed as the following union of combinations of these level sets:
Union(for i varying over the nodes) of (Vx[i] times (product(for j!=i) of the Vn[j])).
So a duplicate complete plan is produced with Nx=Nn and a central composite plan with Nn=1. Intermediate plans for 1<Nn<Nx.
When all involved nodes can return to plan0 the Nx number of levels, the number of observation of the plan is given by p.(Nx).(Nn)^p where p is the number of nodes.

Usage

plan1(bn, ion, Nx=7, Nn=3, repl=0)

Arguments

bn	The original bn.
ion	indicates the node(s) from which values must be fixed (iin or name).
Nx	number of values to extract for expanded nodes.
Nn	number of values to extract for node in the central part.
repl	abs(repl) is the number of replications to provide once the first replication has been done. When the number is negative, TRUE repetition of the first replication are not numerically removed. 0 is equivalent to -1.

Details

Notice that the result can be random, so it is advisable to monitor the pseudo-random seed value.
Multivariate nodes are accepted and their variables are considered as independent nodes.

Value

The resulting data.frame of the values. Columns have got the node names; categoric are factors.

Examples

 rebastaba3k("RESET"); # (only for R checking)
 plan1(rebastaba.bn2, 1);
 plan1(rebastaba.bn2, 2);

rebastaba documentation built on May 2, 2019, 5:24 p.m.