resolve.missing.chain: Solves a chain of missing variables

In samthiriot/gosp.dpp: Pairing of entities to generate structured synthetic populations

Description

For a given incomplete solution, a set of dependant variables to solve, a generation case, and targets defined by the user, tries to identify the best solution (if any).

Usage

resolve.missing.chain(sol, chain, case, nA, nB, nu.A, phi.A, delta.A, nu.B,
  phi.B, delta.B, gamma, verbose = FALSE, tolerance.pd = 1.5e-08,
  tolerance.degree.max = 1.5e-08, tolerance.pij = 1e-06)

Arguments

sol	the current solution (a named list)
chain	the chain to be solved (a list of strings containing variable names)
case	the case to solve
nA	the target population size for A
nB	the target population size for B
nu.A	control for nA: 0 means "respect nA", non-null "adapt it to solve the case"
phi.A	control for frequencies: 0 means "respect the original frequencies as detected in the sample", non-null "adapt it to solve the case"
delta.A	control for degree A: 0 means "respect the input parameters pdi", non-null "adapt them to solve the case"
nu.B	control for nB: 0 means "respect nB", non-null "adapt it to solve the case"
phi.B	control for frequencies: 0 means "respect the original frequencies as detected in the sample", non-null "adapt it to solve the case"
delta.B	control for degree B: 0 means "respect the input parameters pdj", non-null "adapt them to solve the case"
gamma	control for pij: 0 means "respect the matching probabilities pij", non-null "adapt them to solve the case"
verbose	if TRUE, will display detailed information on the console
tolerance.pd	the tolerance for probability distributions
tolerance.degree.max	the tolerance for min/max average degrees
tolerance.pij	the tolerance for pij probabilities

Details

It works by (i) trying to define hypothesis such as "let's try to respect the ideal value for this variable". It then (ii) relies on inference to determine the consequences of this hypothesis, and checks (iii) whether this solution is consistent.

At the end of this process, we might have no solution, one unique solution or several ones. If several solutions are available, the best one is taken, with best being defined as the solution which minimizes the cumulated errors weighted by the user weights.

Value

a list of vectors (the chains) of strings

Author(s)

Samuel Thiriot <[e-mail]>

See Also

propagate.direct for the inference of the consequences of the hypothesis, detect.problems to ensure potential solutions are consistent

samthiriot/gosp.dpp documentation built on May 18, 2019, 3:44 p.m.