############################################################################# 
#                                                                           #                            
#  Copyright: Florian Schuberth                                             #
#  Contact: f.schuberth@utwente.nl                                          #
#                                                                           #
############################################################################# 

# load packages
library(cSEM)

# Specify model without gender.
# gender will be used as grouping variable.
model_Bergami="
# Measurement models
OrgPres <~ cei1 + cei2 + cei3 + cei4 + cei5 + cei6 + cei7 + cei8 
OrgIden =~ ma1 + ma2 + ma3 + ma4 + ma5 + ma6
AffJoy =~ orgcmt1 + orgcmt2 + orgcmt3 + orgcmt7
AffLove  =~ orgcmt5 + orgcmt6 + orgcmt8

# Structural model 
OrgIden ~ OrgPres 
AffLove ~ OrgPres+OrgIden
AffJoy  ~ OrgPres+OrgIden"

# Estimate the model for both gender groups, i.e., gender = 1 and gender = 2, using the .id argument.
# In fact, in the background the dataset is split with regard to gender and 
# the model is estimated twice.
outBergami <- csem(.data = BergamiBagozzi2000,.model = model_Bergami,
                   .disattenuate = T,
                   .PLS_weight_scheme_inner = 'factorial',
                   .tolerance = 1e-6,
                   .resample_method = 'bootstrap',
                   .id='gender')

# Get a summary of the estimation results
summarize(outBergami)

# Conduct MICOM
# Note, configural invariance need to be ensured by the researcher.
# Moreover, MICOM is not suited for testing measurement invariance of latent variables.
outMICOM=testMICOM(outBergami)

# See the MICOM results
outMICOM

# Multigroup analysis using the testMGD function
# By default all parameters are compared across groups
outoverall=testMGD(.object = outBergami,
                   .type_vcv='construct' # which correlation matrix should be compared across groups in the test of Klesel et al. 
                   )

# See the summary results of the MGA.
outoverall

# Get more detailed insights
print(outoverall,.approach_mgd = 'Klesel')
print(outoverall,.approach_mgd = 'Chin')

# If you compare several parameters across groups ensure a sufficient number of permutation runs.
# Moreover, it is recommended to apply a correction for multiple comparison, e.g., Bonferroni
outoverall1=testMGD(.object = outBergami,
                   .type_vcv='construct', # which correlation matrix should be compared across groups using the test of Klesel et al. 
                   .R_permutation = 999,
                   .approach_p_adjust = 'bonferroni')

print(outoverall1,.approach_mgd = 'Chin')


# Test only specific effects, e.g., OrgPres -> OrgIden using the permutation-test of Chin & Dibbern.
# For that purpose you can specify the model and thus the path coefficients in lavaan syntax.
outMGA=testMGD(.object = outBergami,
               .parameters_to_compare = 'OrgIden~OrgPres',
               .approach_mgd = 'Chin')

# See results
outMGA