TopKSignal (version 1.0)

estimateTheta: Estimation of the underlying signal.

Description

The main function for the estimation of the signals informing the ranks is called estimateTheta(). The required parameters are: (1) a rank matrix, (2) the number of bootstrap samples (500 is recommended), (3) a constant for the support variables \(b>0\), default is 0.1, (4) the type of optimization technique: fullLinear, fullQuadratic, restrictedLinear, and restrictedQuadratic (the latter two recommended), (5) the type of bootstrap sampling scheme: classic.bootstrap and poisson.bootstrap (recommended), and (6) the number of cores for parallel computation. Each bootstrap sample is executed on a dedicated CPU core.

Usage

estimateTheta(
  R.input,
  b,
  num.boot,
  solver,
  type,
  bootstrap.type,
  nCore = ((detectCores() - 1))
)

Value

A list with the estimation information obtained:

• estimation - A data frame with the signal estimation and the standard error computed by the bootstrap for each object

• estimatedMatrixNoise - The estimated matrix noise

• time - The execution time of the procedure

• allBootstraps - The signal estimates from all bootstrap iterations

Arguments

R.input

A matrix where the rows represent the objects and the columns the assessors (rankers).

b

The penalization term. The suggested value is 0.1.

num.boot

The number of boostrap samples created from the input ranked matrix. A positive number is expected.

solver

A string that indicates which solver to use. Two options are available, 'gurobi' and 'nloptr'. We recommend to use gurobi for faster computation. Note, a licence is required. Check the corresponding documentation on how to install gurobi.

type

A string that indicates which model to use: four approaches are available: 'restrictedQuadratic', 'fullQuadratic', 'restrictedLinear' and 'fullLinear'.

bootstrap.type

A string that indicates which bootstrap method to use: 'classic.bootstrap' or 'poisson.bootstrap'.

nCore

The number of cores used for computation. Each core is used to calculate the signals from a bootstrap sample. Default number is detectCores() - 1.

Examples

library(TopKSignal)
set.seed(1421)
p = 8
n = 10
input <- generate.rank.matrix(p, n)
rownames(input$R.input) <- c("a","b","c","d","e","f","g","h")
# For the following code Gurobi needs to be installed
if (FALSE) {
estimatedSignal <- estimateTheta(R.input = input$R.input, num.boot = 50, b = 0.1, 
solver = "gurobi", type = "restrictedQuadratic", bootstrap.type = "poisson.bootstrap",nCore = 1)   
}
data(estimatedSignal)
estimatedSignal