Automated model specification search

Usage

doModelSearch(
 .object          = NULL, 
 .pop_size        = 20,
 .n_generations   = 20,
 .prob_mutation   = 0.5,
 .prob_crossover  = 0.8,
 .fbar            = -100000,
 .ms_criterion    = c('bic','aic','hq'),
 .seed            = NULL,
 .only_structural = FALSE
)

Arguments

.object: An R object of class cSEMResults resulting from a call to csem().
.pop_size: Integer. Population size used in the genetic algorithm. Defaults to 20.
.n_generations: Integer. Number of generations used in the genetic algorithm. Defaults to 20.
.prob_mutation: Scalar. Mutation probability used in the genetic algorithm. Defaults to 0.5.
.prob_crossover: Scalar. Crossover probability used in the genetic algorithm. Defaults to 0.8.
.fbar: Integer. Low fitness value that is used to penalize inadmissible models. Defaults to -100000.
.ms_criterion: Character string. A single character string naming the model selection criterion to compute. One of "bic", "aic", or "hq". Defaults to "bic"
.seed: Integer or NULL. The random seed to use. Defaults to NULL in which case an arbitrary seed is chosen. Note that the scope of the seed is limited to the body of the function it is used in. Hence, the global seed will not be altered!
.only_structural: Should the the log-likelihood be based on the structural model? Ignored if .by_equation == TRUE. Defaults to TRUE.

Value

A list of class cSEMModelSearch. The list contains the follwing elements:

$Information: A list of input parameters to doModelSearch
$Results: A list with elements, original_fitness, best_fitness, and best_matrix
$Inputcsem: A list with elements, data, and model, where data comprises the dataset used to estimate the original model and model is a list of cSEMModel lists.

Details

Performs an automated model specification search using a genetic algorithm (Holland 1992) in combination with partial least squares path modeling (PLS-PM). Specifically, the function implements the approach presented in Trinchera et al. (2026) . The model search is limited to the structural model. Thereby it is ensured that the resulting structural model is recursive, i.e., feedback loops and non-recursive structures are prohibited. Moreover, the set of exogenous constructs is taken from the original model provided to doModelSearch() and remains exogenous during the search. If a model violates the restrictions or does not pass verify() during the model search, a penalty fitness value .fbar is assigned to the model. Therefore, ensure that .fbar is sufficiently small. The fitness of the model is determined by the criterion provided to .ms_criterion.

To estimate the found model Inputcsem results can be used, see the example.

References

Holland JH (1992). “Genetic Algorithms.” Scientific American, 267(1), 66–72. doi:10.1038/scientificamerican0792-66 .

Trinchera L, Pietropolli G, Castelli M, Schuberth F (2026). “Automated specification search for composite-based structural equation modeling: a genetic approach.” Computational Statistics & Data Analysis.

Examples

if (FALSE) { # \dontrun{
# Perform model search for a linear model without second-order constructs 
model_Bergami_Bagozzi_Henseler="
 # Measurement models
 OrgPres =~ cei1 + cei2 + cei3 + cei4 + cei5 + cei6 + cei7 + cei8 
 OrgIden =~ ma1 + ma2 + ma3 + ma4 + ma5 + ma6
 AffLove =~ orgcmt1 + orgcmt2 + orgcmt3 + orgcmt7
 AffJoy  =~ orgcmt5 + orgcmt8
 Gender  <~ gender
 
 # Structural model 
 OrgIden ~ OrgPres
 AffLove ~ OrgPres + OrgIden + Gender 
 AffJoy  ~ OrgPres + OrgIden + Gender 
 "
 
 out <- csem(.data = BergamiBagozzi2000, 
             .model = model_Bergami_Bagozzi_Henseler,
             .PLS_weight_scheme_inner = 'factorial',
             .tolerance = 1e-06
 )

 
 outSearch <- doModelSearch(.object = out,
               .pop_size = 20,
               .n_generations = 20,
               .prob_mutation = 0.5,
               .prob_crossover = 0.8,
               .fbar = -100000,
               .ms_criterion = 'bic',
               .seed = 1234) 
 outSearch
 
#  Estimate the found model
 outNew <- csem(.data = outSearch$Inputcsem$data,
                .model = outSearch$Inputcsem$model[[1]])

 # summarize(outNew) 
} # }