numExpSimData-dataset
(Package: clere) :
Performances of 9 methods for dimension reduction on data simulated under the CLERE model
This dataset is a matrix of 200 rows and 28 colums. The columns can be grouped as three blocs of 9 (for each method compared: LASSO, RIDGE, Elastic net [ELNET], Stepwise variable selection [STEP], CLERE, CLERE sparse [CLERE_s], Spike and Slab [SS], AVG method and Pairwise Absolute Clustering and Sparsity [PACS]). Prediction errors (MSE), number of estimated parameters and time (seconds) to fit the data are compared.The 1st 9 (1:9) contain prediction error obtained by 5-fold cross validation using 10 random permutation of the covariate matrix. The 2nd 9 columns (10:18) contain the number of parameters estimated for each method. The 3rd 9 columns are times in seconds measured for fitting each methods. The 28 column is the seed utilized for generating random numbers in these analyses. Each row corresponds to a simulated dataset on which all 9 methods were fitted. For more details, please refer to the package vignette. The R script used to create this dataset is clere/inst/doc/SimulatedDataExample.R.
numExpRealData-dataset
(Package: clere) :
Performances of 9 methods for dimension reduction applied to 2 published real dataset
This data contains two matrices: one for the Prostate dataset (from R package lasso2) and the other for the eyedata dataset (from R package flare). Each matrix has 5 rows and 28 colums. The columns can be grouped as three blocs of 9 (for each method compared: LASSO, RIDGE, Elastic net [ELNET], Stepwise variable selection [STEP], CLERE, CLERE sparse [CLERE_s], Spike and Slab [SS], AVG method and Pairwise Absolute Clustering and Sparsity [PACS]). The 1st 9 (1:9) contain prediction error obtained by 5-fold cross validation using 10 random permutation of the covariate matrix. The 2nd 9 columns (10:18) contain the number of parameters estimated for each method. The 3rd 9 columns are times in seconds measured for fitting each methods. The 28 column is the seed utilized for generating random numbers in these analyses. For more details, please refer the package vignette. The R script used to create this dataset is clere/inst/doc/RealDataExample.R.
This data contains four matrices corresponding to four performance indictors used to compare SEM algorithm and three versions of the MCEM algorithm (MCEMA: with 5 MC interations; MCEMB: with 25 MC iterations and MCEMC: 125 MC iterations) as described in the package vignette. The first matrix Pred contains prediction errors; matrix Bias contains the bias over all model parameters, matrix Time contains execution times for the four methods and matrix Liks the log-likelihood reached by each method. These data were used to generate the Table 1. in the package vignette. For more details, please refer to the package vignette. The R script used to create this dataset is clere/inst/doc/SEM_vs_MCEM_simulations.R.
providing 
.