### Method Compair plots.
for (file in list.files('results', pattern = 'method_compair_[0-9T-]+\\.csv')) {
    # Read simulation result data set.
    ds <- read.csv(file.path('results', file), stringsAsFactors = FALSE)
    # Get dataset groups
    datasets <- unlist(unique(ds["dataset"]))
    # All seperate dataset plots into one file.
    pdf(file.path('results', sub('.csv', '.pdf', file, fixed = TRUE)))
    for (ds.name in datasets) {
        boxplot(ds[ds["dataset"] == ds.name, colnames(ds) != "dataset"],
                main = ds.name, ylab = 'Error', las = 2)
    }
    dev.off()
    # Plot each dataset seperately.
    for (ds.name in datasets) {
        pdf(file.path('results', sub('.csv',
                                     paste0('_', ds.name, '.pdf'),
                      file, fixed = TRUE)))
        boxplot(ds[ds["dataset"] == ds.name, colnames(ds) != "dataset"],
                main = ds.name, ylab = 'Error', las = 2)
        dev.off()
    }
}

### Method Compair plots for parameterized M2.
for (file in list.files('results', 'method_compair_M2_[0-9T-]+\\.csv')) {
    # Read simulation results.
    ds <- read.csv(file.path('results', file), stringsAsFactors = FALSE)
    # Extract M2 dataset parameters.
    pmix <- unique(ds$pmix)
    lambda <- unique(ds$lambda)
    # setup pdf file and plot layout.
    pdf(file.path('results', sub('.csv', '.pdf', file, fixed = TRUE)),
        width = 15, height = 10)
    layout(matrix(seq(length(pmix) * length(lambda)),
                  ncol = length(lambda), byrow = TRUE))
    for (p in pmix) {
        for (l in lambda) {
            boxplot(ds[ds$pmix == p & ds$lambda == l,
                       !(colnames(ds) %in% c("pmix", "lambda"))],
                    main = substitute(expression(paste(
                        lambda, ' = ', l, ', ', p[mix], ' = ', q
                    )), list(l = l, q = p)))
        }
    }
    dev.off()
}

### Analyse predict dimension simulation.
for (file in list.files('results', 'predict_dim_compair_[0-9T-]+\\.csv')) {
    # Read simulation result data set.
    ds <- read.csv(file.path('results', file), stringsAsFactors = FALSE)
    # Over all datasets (including M2 with default parameters)
    true.dim <- ds$true.dim
    dataset <- ds$dataset
    df <- ds[!(colnames(ds) %in% c("true.dim", "dataset"))]
    accuracy <- data.frame(
        all = colMeans(df == true.dim),
        M1 = colMeans(df[dataset == "M1", ] == true.dim[dataset == "M1"]),
        M2 = colMeans(df[dataset == "M2", ] == true.dim[dataset == "M2"]),
        M3 = colMeans(df[dataset == "M3", ] == true.dim[dataset == "M3"]),
        M4 = colMeans(df[dataset == "M4", ] == true.dim[dataset == "M4"]),
        M5 = colMeans(df[dataset == "M5", ] == true.dim[dataset == "M5"]),
        M6 = colMeans(df[dataset == "M6", ] == true.dim[dataset == "M6"]),
        M7 = colMeans(df[dataset == "M7", ] == true.dim[dataset == "M7"])
    )
    cat("Accuracy of simulation:", file, '\n')
    print(accuracy)    
}