controlIsta {lessSEM}

controlIsta

Description

controlIsta

Usage

controlIsta(
  startingValues = "est",
  saveDetails = FALSE,
  L0 = 0.1,
  eta = 2,
  accelerate = TRUE,
  maxIterOut = 10000,
  maxIterIn = 1000,
  breakOuter = 1e-08,
  convCritInner = 1,
  sigma = 0.1,
  stepSizeInheritance = ifelse(accelerate, 1, 3),
  verbose = 0,
  nCores = 1
)

Arguments

startingValues

option to provide initial starting values. Only used for the first lambda. Three options are supported. Setting to "est" will use the estimates from the lavaan model object. Setting to "start" will use the starting values of the lavaan model. Finally, a labeled vector with parameter values can be passed to the function which will then be used as starting values.

saveDetails

when set to TRUE, additional details about the individual models are save. Currently, this are the implied means and covariances. Note: This may take a lot of memory!

L0

L0 controls the step size used in the first iteration

eta

eta controls by how much the step size changes in the inner iterations with (eta^i)*L, where i is the inner iteration

accelerate

boolean: Should the acceleration outlined in Parikh, N., & Boyd, S. (2013). Proximal Algorithms. Foundations and Trends in Optimization, 1(3), 123–231., p. 152 be used?

maxIterOut

maximal number of outer iterations

maxIterIn

maximal number of inner iterations

breakOuter

change in fit required to break the outer iteration. Note: The value will be multiplied internally with sample size N as the -2log-Likelihood depends directly on the sample size

convCritInner

this is related to the inner breaking condition. 0 = ista, as presented by Beck & Teboulle (2009); see Remark 3.1 on p. 191 (ISTA with backtracking) 1 = gist, as presented by Gong et al. (2013) (Equation 3)

sigma

sigma in (0,1) is used by the gist convergence criterion. larger sigma enforce larger improvement in fit

stepSizeInheritance

how should step sizes be carried forward from iteration to iteration? 0 = resets the step size to L0 in each iteration 1 = takes the previous step size as initial value for the next iteration 3 = Barzilai-Borwein procedure 4 = Barzilai-Borwein procedure, but sometimes resets the step size; this can help when the optimizer is caught in a bad spot.

verbose

if set to a value > 0, the fit every "verbose" iterations is printed.

nCores

number of core to use. Multi-core support is provided by RcppParallel and only supported for SEM, not for general purpose optimization.

Value

object of class controlIsta

Examples

control <- controlIsta()