nmr_pca_outliers_robust: Outlier detection through robust PCA
In AlpsNMR: Automated spectraL Processing System for NMR

Description Usage Arguments Value See Also Examples

Outlier detection through robust PCA

1	nmr_pca_outliers_robust(nmr_dataset, ncomp = 5)

nmr_dataset

An nmr_dataset_1D object

ncomp

Number of rPCA components to use

We have observed that the statistical test used as a threshold for outlier detection usually flags as outliers too many samples, due possibly to a lack of gaussianity

As a workaround, a heuristic method has been implemented: We know that in the Q residuals vs T scores plot from nmr_pca_outliers_plot() outliers are on the right or on the top of the plot, and quite separated from non-outlier samples.

To determine the critical value, both for Q and T, we find the biggest gap between samples in the plot and use as critical value the center of the gap.

This approach seems to work well when there are outliers, but it fails when there isn't any outlier. For that case, the gap would be placed anywhere and that is not desirable as many samples would be incorrectly flagged. The second assumption that we use is that no more than 10\ the samples may pass our critical value. If more than 10\ pass the critical value, then we assume that our heuristics are not reasonable and we don't set any critical limit.

A list similar to nmr_pca_outliers

Other PCA related functions: nmr_pca_build_model(), nmr_pca_outliers_filter(), nmr_pca_outliers_plot(), nmr_pca_outliers(), nmr_pca_plots

Other outlier detection functions: Pipelines, nmr_pca_outliers_filter(), nmr_pca_outliers_plot(), nmr_pca_outliers()

Other nmr_dataset_1D functions: [.nmr_dataset_1D(), computes_peak_width_ppm(), file_lister(), files_to_rDolphin(), format.nmr_dataset_1D(), is.nmr_dataset_1D(), load_and_save_functions, new_nmr_dataset_1D(), nmr_align_find_ref(), nmr_baseline_removal(), nmr_baseline_threshold(), nmr_exclude_region(), nmr_integrate_regions(), nmr_interpolate_1D(), nmr_meta_add(), nmr_meta_export(), nmr_meta_get_column(), nmr_meta_get(), nmr_normalize(), nmr_pca_build_model(), nmr_pca_outliers_filter(), nmr_pca_outliers_plot(), nmr_pca_outliers(), nmr_ppm_resolution(), plot.nmr_dataset_1D(), plot_webgl(), print.nmr_dataset_1D(), rdCV_PLS_RF_ML(), rdCV_PLS_RF(), save_files_to_rDolphin(), to_ChemoSpec(), validate_nmr_dataset_peak_table(), validate_nmr_dataset()

dir_to_demo_dataset <- system.file("dataset-demo", package = "AlpsNMR")
dataset <- nmr_read_samples_dir(dir_to_demo_dataset)
dataset_1D <- nmr_interpolate_1D(dataset, axis = c(min = -0.5, max = 10, by = 2.3E-4))
outliers_info <- nmr_pca_outliers_robust(dataset_1D)