compute.slices: Compute the objective function on a grid of params and show...

In tlamadon/mopt: Parallel Derivative-free Moment Estimation

Description

Will vary each parameter independently in a chosen range and report the value of the resulting simulated moments in relation to the moments in the data. Can be used to construct a heuristic identification argument. Basically it can be seen which parameter affects which dimension of the model output, i.e. which simulated moment.

Usage

compute.slices(mcf, ns = 30, pad = 0.1, file = "est.slices.RData")

Arguments

mcf	object of class mopt
ns	number of points in each dimension to evaluate
pad	from bounds of parameter ranges. e.g. p in [0,1], avoid 0 and 1 with pad>0.
file	/path/to/your/file

Value

list with info and a data.frame slices summarizing all information of the exercise: parameter values, simulated moments, data moments. Input to plot.slices(slices).

Examples

# generate slices for a model
require(mopt)

# let's take a dummy objective function
MOPT_OBJ_FUNC <- objfc.norm2(c(0,0),ns=2000)

# starting parameters
p <- list(x1=0.5,x2=0.5)
MOPT_OBJ_FUNC(p)

# then we want to setup the mopt
mcf                  = mopt_config(p)
mcf$wd               = getwd()
mcf$params_to_sample = c('x1','x2')
mcf$moments_to_use   = c('m1','m2')
mcf$mode             = 'multicore'
mcf$algo             = algo.bgp


# set the parameter bounds
mcf <- mcf + 
  samplep('x1',-1,1) +
  samplep('x2',-1,1)

# adding data moment values
mcf <- mcf + datamoments(c('m1','m2'),
                           c(0,0),
                           c(0.1,0.1))

# prepare to run with OpenMP
require(parallel)
options(mc.cores = detectCores())

# finalize the preparation
mcf <- prepare.mopt_config(mcf)

# compute slices and generate plots
res <- compute.slices(mcf,ns=30,pad=0.1)

tlamadon/mopt documentation built on May 31, 2019, 3:48 p.m.