cSEM (version 0.1.0)

predict: Predict indicator scores

Description

Predict the indicator scores of endogenous constructs.

Usage

predict(
 .object               = NULL,
 .benchmark            = c("lm", "unit", "PLS-PM", "GSCA", "PCA", "MAXVAR"),
 .cv_folds             = 10,
 .handle_inadmissibles = c("stop", "ignore", "set_NA"),
 .only_common_factors  = TRUE,
 .r                    = 10,
 .test_data            = NULL
 )

Arguments

.object

An R object of class cSEMResults resulting from a call to csem().

.benchmark

Character string. The procedure to obtain benchmark predictions. One of "lm", "unit", "PLS-PM", "GSCA", "PCA", or "MAXVAR". Default to "lm".

.cv_folds

Integer. The number of cross-validation folds to use. Setting .cv_folds to N (the number of observations) produces leave-one-out cross-validation samples. Defaults to 10.

.handle_inadmissibles

Character string. How should inadmissible results be treated? One of "stop", "ignore", or "set_NA". If "stop", predict() will stop immediatly if estimation yields an inadmissible result. For "ignore" all results are returned even if all or some of the estimates yielded inadmissible results. For "set_NA" predictions based on inadmissible parameter estimates are set to NA.

.only_common_factors

Logical. Should only indicator scores for concepts modeled as common factors be predicted? Defaults to TRUE.

.r

Integer. The number of repetitions to use. Defaults to 10.

.test_data

A matrix of test data with the same column names as the training data.

Value

An object of class cSEMPredict with print and plot methods. Technically, cSEMPredict is a named list containing the following list elements:

$Actual

A matrix of the actual values/indicator scores of the endogenous constructs.

$Prediction_target

A matrix of the predicted indicator scores of the endogenous constructs based on the target model. Target refers to

$Residuals_target

A matrix of the residuals indicator scores of the endogenous constructs based on the target model.

$Residuals_lm

A matrix of the residuals indicator scores of the endogenous constructs based on a linear model in which the indicator scores of endogenous constructs are predicted by exogenous indicator scores. This serves as a benchmark for comparisons.

$Prediction_metrics

A data frame containing the predictions metrics MAE, RMSE, and Q2_predict.

$Information

A list with elements Target, Benchmark, Number_of_observations_training, Number_of_observations_test, Number_of_folds, Number_of_repetitions, and Handle_inadmissibles.

Details

Predict uses the procedure introduced by Shmueli2016;textualcSEM in the context of PLS (commonly called: "PLSPredict" Shmueli2019cSEM). Predict uses k-fold cross-validation to randomly split the data into training and test data and subsequently predicts the relevant values in the test data based on the model parameter estimates obtained using the training data. The number of cross-validation folds is 10 by default but may be changed using the .cv_folds argument. By default, the procedure is repeated .r = 10 times to avoid irregularities due to a particular split. See Shmueli2019;textualcSEM for details.

Alternatively, users may supply a matrix of .test_data with the same column names as those in the data used to obtain .object (the training data). In this case, arguments .cv_folds and .r are ignored and predict uses the estimated coefficients from .object to predict the values in the columns of .test_data.

In Shmueli2016;textualcSEM PLS-based predictions for indicator i are compared to the predictions based on a multiple regression of indicator i on all available exogenous indicators (.benchmark = "lm") and a simple mean-based prediction summarized in the Q2_predict metric. predict() is more general in that is allows users to compare the predictions based on a so-called target model/specificiation to predictions based on an alternative benchmark. Available benchmarks include predictions based on a linear model, PLS-PM weights, unit weights (i.e. sum scores), GSCA weights, PCA weights, and MAXVAR weights.

By default, only the indicator scores of constructs modeled as common factors are predicted (.only_common_factors = TRUE). While technically possible, prediction for constructs modeled as composites is conceptually difficult since composites are by design build by their indicators, i.e., composites are not though of as being predictive of their indicators.

Each estimation run is checked for admissibility using verify(). If the estimation yields inadmissible results, predict() stops with an error ("stop"). Users may choose to "ignore" inadmissible results or to simply set predictions to NA ("set_NA") for the particular run that failed.

References

See Also

csem, cSEMResults

Examples

Run this code
# NOT RUN {
### Anime example taken from https://github.com/ISS-Analytics/pls-predict

# Load data
data(Anime) # data is similar to the Anime.csv found on 
            # https://github.com/ISS-Analytics/pls-predict but with irrelevant
            # columns removed

# Split into training and data the same way as it is done on 
# https://github.com/ISS-Analytics/pls-predict
set.seed(123)

index     <- sample.int(dim(Anime)[1], 83, replace = FALSE)
dat_train <- Anime[-index, ]
dat_test  <- Anime[index, ]

# Specify model
model <- "
# Structural model

ApproachAvoidance ~ PerceivedVisualComplexity + Arousal

# Measurement/composite model

ApproachAvoidance         =~ AA0 + AA1 + AA2 + AA3
PerceivedVisualComplexity <~ VX0 + VX1 + VX2 + VX3 + VX4
Arousal                   <~ Aro1 + Aro2 + Aro3 + Aro4
"

# Estimate (replicating the results of the `simplePLS()` function)
res <- csem(dat_train, 
            model, 
            .disattenuate = FALSE, # original PLS
            .iter_max = 300, 
            .tolerance = 1e-07, 
            .PLS_weight_scheme_inner = "factorial"
)

# Predict using a user-supplied training data set
pp <- predict(res, .test_data = dat_test)
pp$Predictions_target[1:6, ]
pp

### Compute prediction metrics  ------------------------------------------------
res2 <- csem(Anime, # whole data set
            model, 
            .disattenuate = FALSE, # original PLS
            .iter_max = 300, 
            .tolerance = 1e-07, 
            .PLS_weight_scheme_inner = "factorial"
)

# Predict using 10-fold cross-validation with 5 repetitions
pp2 <- predict(res, .benchmark = "lm")
pp2
# }

Run the code above in your browser using DataCamp Workspace