Package {metabodeconplus}

metabodeconplus: Deconvolution, Alignment and Model Fitting of 1d NMR Spectra

Description

An integrated framework for deconvolution, alignment and postprocessing of 1-dimensional (1d) nuclear magnetic resonance (NMR) spectra, extended with end-to-end model fitting that turns the resulting matrix of aligned signal integrals into classification models. The deconvolution part uses the algorithm described in Koh et al. (2009) doi:10.1016/j.jmr.2009.09.003. The alignment part is based on functions from the 'speaq' package, described in Beirnaert et al. (2018) doi:10.1371/journal.pcbi.1006018 and Vu et al. (2011) doi:10.1186/1471-2105-12-405. A detailed description and evaluation of an early version of the package can be found in Haeckl et al. (2021) doi:10.3390/metabo11070452. 'metabodeconplus' is the actively developed successor to the 'metabodecon' package and introduces backwards-incompatible API changes.

Author(s)

Maintainer: Tobias Schmidt tobias.schmidt331@gmail.com [copyright holder]

Authors:

• Martina Haeckl martina.haeckl@stud.uni-regensburg.de [copyright holder]

• Wolfram Gronwald wolfram.gronwald@klinik.uni-regensburg.de [copyright holder]

Other contributors:

• Yanren Linda Hu yanren-linda.hu@klinik.uni-regensburg.de [contributor]

See Also

Useful links:

• https://github.com/spang-lab/metabodeconplus/

• https://spang-lab.github.io/metabodeconplus/

• Report bugs at https://github.com/spang-lab/metabodeconplus/issues

Get URL of Metabodecon "Get Started" Page

Description

get_started and aaa_Get_Started both return (and optionally open) the URL of the "Get Started" page of the metabodeconplus documentation. The aaa_Get_Started version exists, because functions are listed alphabetically in the reference manual and we want get_started to be shown at the top of the list (i.e., it needs to start with an 'a').

Usage

aaa_Get_Started(open_browser = interactive())

get_started(open_browser = interactive())

Arguments

Value

A character string containing the URL of the "Get Started" page.

Author(s)

2024-2025 Tobias Schmidt: initial version.

Examples

get_started(open_browser = FALSE)
get_started()

Align deconvoluted spectra

Description

Aligns peaks across a set of deconvoluted spectra by chaining two stages:

1. CluPA (clupa()) shifts peak centers continuously toward a reference using hierarchical-clustering FFT segment shifts (Beirnaert et al. 2018, Vu et al. 2011). Adds x0al and pcial (post-CluPA center and column index) to each peak; original x0, A, lambda, pcide are preserved.

2. Reference snapping (snap_to_ref()) records, for each peak, the nearest reference column within maxCombine as pcisn / x0sn. Peaks farther than maxCombine from every reference column get pcisn = NA / x0sn = NA. No peaks are dropped and amplitudes are not summed here — collisions on the same pcisn are aggregated downstream by si_mat().

All spectra in x must already live on the same chemical-shift grid (identical ⁠$cs⁠ vector across spectra). Call harmonize_grid() upstream if your inputs come from different acquisitions with slight calibration offsets.

Usage

align(
  x,
  y = NULL,
  ref = NULL,
  maxShift = 50,
  maxCombine = 0,
  verbose = TRUE,
  nworkers = 1,
  full = TRUE,
  use_speaq = FALSE,
  gap_tol = NULL
)

Arguments

Value

An object of class aligns.

Author(s)

2024-2026 Tobias Schmidt: initial version.

Examples

decons <- deconvolute(sim[1:5], sfr=c(3.55, 3.35), verbose=FALSE)
aligned <- align(decons, maxShift=50, maxCombine=20, verbose=FALSE)

Alignment building blocks

Description

Pluggable alignment stages used by align() and the align_fun argument of fit_mdm() / benchmark().

• clupa(): CluPA — hierarchical-clustering peak alignment (recursive FFT segment shifts, Beirnaert et al. 2018, Vu et al. 2011). Operates on the Lorentz reconstruction sit$sup already attached to each spectrum by deconvolution.

• snap_to_ref(): Reference snapping — snap each peak to the nearest reference column within maxCombine.

All these functions require every input spectrum to share the same ⁠$cs⁠ grid; an explicit stop() is raised otherwise. Call harmonize_grid() upstream to enforce that invariant.

snap_to_ref() applies the snapping step on its own: for each peak in each spectrum, finds the nearest reference column on the shared cs grid and records that column as pcisn (and its ppm value as x0sn). Peaks farther than maxCombine columns from every reference column get pcisn = NA / x0sn = NA. Original x0, x0al, A, lambda, pcide and pcial are preserved — snapping only adds the snapped fields. Collisions on the same pcisn column are not merged here; si_mat() sums their areas when rasterising the feature matrix. sit$supal is cleared because the post-snap superposition would need recomputing.

Usage

clupa(
  x,
  y = NULL,
  ref = NULL,
  maxShift = 50,
  verbose = TRUE,
  nworkers = 1,
  full = TRUE,
  use_speaq = FALSE,
  gap_tol = NULL
)

snap_to_ref(x, ref = NULL, maxCombine = 20, ...)

Arguments

Value

An object of class aligns.

Examples

decons <- deconvolute(sim[1:5], sfr=c(3.55, 3.35), verbose=FALSE)
aligned <- clupa(decons, maxShift=50, verbose=FALSE)  # CluPA stage
snapped <- snap_to_ref(aligned, maxCombine=20)        # reference snapping

Bin a spectra-like object into a feature matrix

Description

Bins the per-spectrum signal vector left-to-right into chunks of maxCombine chemical-shift columns and returns the per-bin sums as a feature matrix. Columns whose chemical-shift falls inside any igrs interval are removed before binning.

Accepts three input types:

• spectra: uses x[[i]]$si directly.

• decons2: uses x[[i]]$sit$sup (smoothed reconstruction).

• aligns: builds a sparse vector from lcpar$pcial / lcpar$A * pi, then bins.

Suitable as the feat_fun argument of fit_mdm() for binning baselines.

Usage

bin(x, maxCombine = 128, igrs = list(), peakPos = NULL, ...)

Arguments

Value

A numeric matrix with one row per spectrum and one column per bin.

Examples

decons <- deconvolute(sim[1:3], sfr=c(3.55, 3.35), verbose=FALSE)
X <- bin(decons, maxCombine=50)

700-bin Zacharias 2013 feature matrix

Description

Builds a feature matrix on the fixed 700-bin grid of Zacharias (2013): 300 bins covering 6.5-9.5 ppm + 400 bins covering 0.5-4.5 ppm, both at 0.01 ppm width. The water region (4.5-6.5 ppm) is excluded. Bins are ordered high-to-low — column 1 covers (9.49, 9.50) ppm, column 700 covers (0.50, 0.51) ppm.

Suitable as the feat_fun argument of fit_mdm(). Per-spectrum dispatch:

• If lcpar is empty (raw spectra): bin ⁠$si⁠ directly.

• If lcpar is non-empty (deconvoluted / aligned spectra): reconstruct si_hat = lorentz_sup(cs, lcpar) on the spectrum's own cs grid using x0al when present (else x0), then bin the reconstruction.

Always returns all 700 columns; maxCombine, igrs and peakPos are accepted for feat_fun protocol compatibility but ignored (the bin layout is hardcoded).

Usage

bin700(x, maxCombine = 0L, igrs = list(), peakPos = NULL, ...)

Arguments

Value

A numeric matrix with one row per spectrum and 700 columns of bin sums. Column names are sprintf("%.4f", bin_midpoint); row names are spectrum names.

Author(s)

2026 Tobias Schmidt: initial version.

Examples

# `sim` spans only ~3.3-3.6 ppm, so most of the 700 fixed bins are 0.
X <- bin700(sim[1:3])

Check Rust Backend Availability

Description

check_mdrb() returns a boolean indicating whether a suitable version of the metabodeconplus Rust backend mdrb is currently installed. The Rust backend is entirely optional; metabodeconplus's pure-R backend is the default and always available.

Usage

check_mdrb(stop_on_fail = FALSE)

Arguments

Value

check_mdrb() returns TRUE if a suitable version of mdrb is installed, else FALSE.

Author(s)

2024-2025 Tobias Schmidt: initial version.

Examples

check_mdrb()

Convert from unit A to unit B

Description

Converts positions/widths from unit A to unit B. If the direction of units A and B is reversed, the width's sign will be reversed as well. To keep widths strictly positive, wrap the result with abs().

Usage

convert_pos(xa, ya, yb)

convert_width(xa, ya, yb)

Arguments

Value

A numeric vector of values converted from unit A to unit B.

Author(s)

2024-2025 Tobias Schmidt: initial version.

Examples

ya <- c(244, 246, 248, 250, 252)
yb <- c(15, 10, 5, 0, -5)
convert_width(c(2, 4, 8), ya, yb)
convert_pos(c(247, 249), ya, yb)

Return path to metabodeconplus's data directory

Description

Returns the path to the directory where download_example_datasets() stores metabodeconplus's example data sets or any file within that directory. By default this directory is a subdirectory of R's temporary session directory. If persistent is set to TRUE, the directory equals the data directory returned by tools::R_user_dir() instead.

Usage

datadir(file = NULL, warn = TRUE, persistent = NULL)

Arguments

Details

The decision to use a temporary data dir as default and a persistent one only optionally was made to conform to CRAN package policies, which state that:

Packages should not write in the user's home filespace (including clipboards), nor anywhere else on the file system apart from the R session's temporary directory [...] Limited exceptions may be allowed in interactive sessions if the package obtains confirmation from the user. For R version 4.0 or later [...] packages may store user-specific data, configuration and cache files in their respective user directories obtained from tools::R_user_dir() [...].

Source: cran.r-project.org/web/packages/policies.

Value

Path to the data directory or a file within it.

Author(s)

2024-2025 Tobias Schmidt: initial version.

See Also

download_example_datasets(), datadir_persistent(), datadir_temp()

Examples

# Get temporary datadir and persistent datadir
datadir(persistent = FALSE, warn = FALSE)
datadir(persistent = TRUE,  warn = FALSE)

# Get persistent datadir if existing else temp datadir. Set `warn = TRUE`
# to raise a warning if none of the directories exist yet.
datadir(warn = FALSE)
if (interactive()) datadir()

# Get PERSISTENT_DATADIR/bruker if existing else TEMP_DATADIR/bruker
datadir(file = "bruker/urine", warn = FALSE)

Persistent Data Directory

Description

Returns the path to the persistent data directory where metabodeconplus's data sets are stored. This directory equals the data directory returned by tools::R_user_dir() plus additional path normalization.

Usage

datadir_persistent()

Value

Path to the persistent data directory.

Author(s)

2024-2025 Tobias Schmidt: initial version.

See Also

datadir(), datadir_temp()

Examples

datadir_persistent()

Temporary Data Directory

Description

Returns the path to the temporary data directory where metabodeconplus's data sets are stored. This directory equals subdirectory 'data' of metabodeconplus's temporary session directory tmpdir() plus additional path normalization.

Usage

datadir_temp()

Value

Returns the path to the temporary data directory.

Author(s)

2024-2025 Tobias Schmidt: initial version.

See Also

tmpdir(), datadir(), datadir_persistent()

Examples

datadir_temp()

Deconvolute one or more NMR spectra

Description

Deconvolutes NMR spectra by modeling each detected signal within a spectrum as Lorentz Curve.

Returns the default grid of ⁠(nfit, smit, smws, delta)⁠ combinations used by deconvolute() when npmax >= 1. Useful as the deg argument to fit_mdm().

Usage

deconvolute(
  x,
  nfit = 3,
  smit = 2,
  smws = 5,
  delta = 6.4,
  npmax = 0,
  sfr = NULL,
  igrs = list(),
  use_rust = FALSE,
  verbose = TRUE,
  nworkers = 1
)

get_deg(conf = "default")

Arguments

Details

First, an automated curvature based signal selection is performed. Each signal is represented by 3 data points to allow the determination of initial Lorentz curves. These Lorentz curves are then iteratively adjusted to optimally approximate the measured spectrum.

Value

A 'decon2' object as described in metabodeconplus-classes.

A data frame with columns nfit, smit, smws, delta.

Author(s)

2024-2025 Tobias Schmidt: initial version.

Examples

## Deconvolute a single spectrum
spectrum <- sim[[1]]
decon <- deconvolute(spectrum)

## Read multiple spectra from disk and deconvolute at once
spectra_dir <- metabodeconplus_file("sim_subset")
spectra <- read_spectra(spectra_dir)
decons <- deconvolute(spectra, sfr = c(3.55,3.35))
get_deg()

Download metabodeconplus Example Datasets

Description

Downloads example datasets that can be used to test the functionality of the metabodeconplus package. These datasets are not included in the package by default due to size constraints. The datasets are downloaded as zip file and extracted automatically, unless extraction is disabled by the user.

Usage

download_example_datasets(
  dst_dir = NULL,
  extract = TRUE,
  persistent = NULL,
  overwrite = FALSE,
  silent = FALSE
)

Arguments

Value

The path to the downloaded (and possibly extracted) datasets.

Author(s)

2024-2025 Tobias Schmidt: initial version.

See Also

datadir()

Examples

if (interactive()) {
     zip <- download_example_datasets(extract = FALSE, persistent = FALSE)
     dir <- download_example_datasets(extract = TRUE)
}

Draw Spectrum

Description

Draws a single spectrum. Internally used by plot_spectrum(), which is usually the recommended way to plot spectra. For usage examples see test/testthat/test-draw_spectrum.R.

Usage

draw_spectrum(
  obj,
  foc_rgn = NULL,
  foc_frac = NULL,
  foc_only = TRUE,
  add = FALSE,
  fig_rgn = NULL,
  main = NULL,
  show = TRUE,
  show_d2 = FALSE,
  truepar = NULL,
  mar = c(4.1, 5.1, 1.1, 1.1),
  sf_vert = "auto",
  si_line = list(),
  sm_line = list(),
  sp_line = list(),
  d2_line = list(),
  al_line = list(),
  lc_lines = list(),
  tp_lines = list(),
  al_lines = list(),
  cent_pts = list(),
  bord_pts = list(),
  norm_pts = list(),
  tp_pts = list(),
  fp_pts = list(),
  miss_pts = list(),
  bg_rect = list(),
  foc_rect = list(),
  lc_rects = list(),
  tp_rects = list(),
  bt_axis = list(),
  lt_axis = list(),
  tp_axis = list(),
  rt_axis = list(),
  bt_text = list(),
  lt_text = list(),
  tp_text = list(),
  rt_text = list(),
  tp_verts = list(),
  lc_verts = list(),
  al_verts = list(),
  ze_hline = list(),
  al_arrows = list(),
  lgd = list()
)

Arguments

Details

Parameters bt_axis, lt_axis, tp_axis and rt_axis all support option n and digits, where n = 5 means "Draw 5 tickmarks over the full axis range" and digits = 3 means "round the label shown beside each tickmark to 3 digits". If n or digits is omitted, a suitable value is chosen automatically. Providing a vector of digits causes each digit to be tried until a digit is encountered that results in n unique labels. Example:

Assume we have n = 4 and the corresponding calculated tickmark positions are: 1.02421, 1.02542, 1.02663 and 1.02784. If we provide digits = 1:5, the following representations are tried:

In the above example the process would stop at digit = 3, because at this point we have n = 4 unique labels (1.024, 1.025, 1.027 and 1.028).

Value

NULL. Called for side effect of plotting.

Author(s)

2024-2025 Tobias Schmidt: initial version.

Examples

decon <- deconvolute(sim[[1]], sfr = c(3.55, 3.35))
draw_spectrum(obj = decon)
draw_spectrum(obj = decon, lgd = list(x = "top", bg = NA))
draw_spectrum(obj = decon, foc_rgn = c(3.45, 3.37))
draw_spectrum(obj = decon, add = FALSE, lgd = FALSE,
              fig = c(.2, .8, .2, .4), mar = c( 0,  0,  0,  0))
draw_spectrum(obj = decon, add = TRUE, lgd = FALSE,
              fig = c(0.2, 0.8, 0.6, 0.8), mar = c(0, 0, 0, 0))
draw_spectrum(obj = decon, lc_lines = NULL, lc_rects = NULL, foc_only = FALSE)

Evaluate an expression with predefined global state

Description

Evaluates an expression with a predefined global state, including the:

• working directory (set via base::setwd())

• global options (set via base::options())

• graphical parameters (set via graphics::par())

In addition to that, evalwith allows to:

• Redirect or capture the output and/or message stream via base::sink()

• Measure the runtime of the evaluated expression via base::system.time()

• Creating a temporary test directory (inside tmpdir()) and populating it with input files according to inputs

• Predefine answers for calls to base::readline() happening during evaluation of expr

• Caching the result of the expression

All changes to the global state are reverted after the expression has been evaluated.

Usage

evalwith(
  expr,
  testdir = NULL,
  answers = NULL,
  output = NULL,
  message = NULL,
  plot = NULL,
  datadir_temp = c("default", "missing", "empty", "filled")[1],
  datadir_persistent = c("default", "missing", "empty", "filled")[1],
  inputs = character(),
  opts = NULL,
  pars = NULL,
  cache = FALSE,
  overwrite = FALSE
)

Arguments

Details

The datadir_temp and datadir_persistent arguments accept values "missing", "filled" and "empty". Setting a value unequal NULL causes the functions datadir_temp() and/or datadir_persistent() to be replaced with mock functions pointing to fake directories. Functions depending on these functions will then use the fake directories instead of the real ones. When set to "missing" the returned mock directory does not exist. When set to "empty" it exists and is guaranteed to be empty. When set to "filled", it is populated with example datasets.

Attention: the mocked functions, i.e. datadir_temp() and datadir_persistent() cannot be used directly inside expr when called via devtools::test(). I'm not sure why, but it seems as if devtools and/or testthat have their own copies of the functions which are used when the expression is evaluated.

Value

A list containing with following elements:

• rv: The return value of the expression.

• runtime: The "elapsed" runtime of the expression in seconds. Measured with base::system.time().

• output: The captured output.

• message: The captured messages.

• plot: The path to the saved plot.

• testdir: The path to the test directory.

• inputs: The paths to the copied input files.

Author(s)

2024-2025 Tobias Schmidt: initial version.

Examples

x1 <- evalwith(output = "captured", cat("Helloworld\n"))
str(x1)

x2 <- evalwith(datadir_persistent = "missing", message = "captured", datadir())
str(x2)

x3 <- evalwith(testdir = "dummy", inputs = "bruker/urine/urine_1", dir())
str(x3)

x4 <- evalwith(Sys.sleep(0.02))
str(x4)

Retrieve directory path of an example dataset

Description

Returns the path to the directory storing the example files shipped with metabodeconplus.

Deprecated since metabodeconplus v1.2.0. Please use datadir() instead. See examples below for usage.

Usage

get_data_dir(
  dataset_name = c("", "blood", "test", "urine", "aki"),
  warn = TRUE
)

Arguments

Value

Path to the directory storing the example files.

Author(s)

2024-2025 Tobias Schmidt: initial version.

See Also

download_example_datasets()

Examples

x <- get_data_dir("urine")                     # Deprecated
y <- datadir("example_datasets/bruker/urine")  # Preferred
cat(x, y, sep = "\n")

Harmonize a corpus of spectra onto a shared chemical-shift grid

Description

Pre-aligns every spectrum in x to a single shared chemical-shift grid by integer-datapoint shifting. After the call, every spectrum's ⁠$cs⁠ is bit-identical to the chosen target grid, so downstream code (deconvolution, clupa(), snap_to_ref(), feature-matrix builders) can index everyone by datapoint and treat cs as a single shared variable.

All input spectra must share the same point count and the same ppm width (calibration offsets are allowed, frequency-domain resolution differences are not). For each spectrum the integer offset from the target grid is computed in datapoints, ⁠$si⁠ is rolled by that many positions, the vacated edge is filled with pad (default 0), and ⁠$cs⁠ is replaced by the target grid.

This kills the per-spectrum absolute-calibration offset between acquisitions (typically a constant shift in ppm coming from spectrometer reference setup) so that downstream CluPA only has to correct the remaining sub-datapoint residual + the random chemical-shift drift from sample composition.

Sub-datapoint residual: at most ±0.5 datapoint per spectrum, which is two orders of magnitude smaller than typical Lorentzian peak widths (lambda ~ 1e-3 to 1e-2 ppm vs. spacing ~ 1e-4 ppm), so the rounding error is irrelevant for any downstream fit.

Edge handling: a spectrum that needs to shift right by k datapoints loses k datapoints from one end and gains k pad values on the other. For typical NMR spectra the edges are noise, so zero-padding is equivalent to dropping the noise — no real signal is harmed. For large absolute shifts (tens of datapoints) you can verify the discarded region is noise by inspecting ⁠$cs⁠ against the metabolite range of interest.

Usage

harmonize_grid(x, target = "median", pad = 0)

Arguments

Value

A spectra object with every spectrum's ⁠$cs⁠ replaced by the target grid and ⁠$si⁠ shifted accordingly. All other fields (⁠$meta⁠, ⁠$lcpar⁠, ⁠$sit⁠, ...) are preserved unchanged.

Author(s)

2026 Tobias Schmidt: initial version.

Examples

# `sim` already shares one cs grid, so harmonization is a no-op here;
# on a corpus from different acquisitions it snaps them to a common grid.
x <- harmonize_grid(sim)
all(sapply(x, function(s) identical(s$cs, x[[1]]$cs)))  # TRUE

Show head and tail rows of a matrix-like object

Description

Returns the first and last n rows of a matrix or data frame. If the input has fewer than 2*n rows, overlapping rows are returned only once.

Usage

headtail(x, n = 6)

Arguments

Value

A subset of x containing head and tail rows.

Author(s)

2024-2026 Tobias Schmidt: initial version.

Examples

x <- matrix(seq_len(30), nrow = 10)
headtail(x, n = 2)

Plot Spectra Heatmap

Description

Plot a set of spectra as a heatmap. Each row corresponds to one spectrum, each column to a chemical-shift datapoint, and the signal intensity is color-coded.

If the spectra were simulated (i.e. carry a simpar element in meta), the true peaks are highlighted with thick rectangles spanning x0 +- lambda. If the spectra have additionally been deconvoluted, the rectangles are colored according to whether each peak was correctly identified (green), missed (yellow) or wrongly identified (red, drawn at the position of the deconvoluted peak).

Usage

heat_spectra(
  objs,
  foc_rgn = NULL,
  what = NULL,
  cols = NULL,
  xlab = "Chemical Shift [ppm]",
  ylab = "Spectrum",
  mar = c(4.1, 2.1, 1.1, 0.5),
  y = NULL,
  y_cols = NULL,
  true_x0 = NULL,
  true_col = "darkgreen",
  true_tol = NULL,
  scale_cols = FALSE,
  cex_names = 0.8,
  xaxis_side = 1,
  col_scores = NULL,
  col_sep = NULL,
  row_sep = NULL,
  main = NULL,
  names = NULL,
  ref = NULL,
  ref_col = "red",
  ref_lwd = NULL,
  sparse = FALSE,
  xaxt = "s"
)

Arguments

Value

NULL. Called for side effect of plotting.

Author(s)

2024-2026 Tobias Schmidt: initial version.

Examples

obj <- deconvolute(sim[1:4], sfr = c(3.55, 3.35))
heat_spectra(obj)
heat_spectra(obj, foc_rgn = c(3.55, 3.35))

Create a Spectrum Object

Description

Creates a spectrum object from the provided signal intensities, frequencies and chemical shifts.

Usage

make_spectrum(
  si,
  cs_max,
  cs_width,
  fq_ref,
  fq_width = NULL,
  force = FALSE,
  silent = FALSE,
  name = NULL,
  path = NULL,
  type = NULL,
  simpar = NULL,
  mfs = NULL
)

Arguments

Value

A spectrum object as described in metabodeconplus-classes.

Author(s)

2024-2025 Tobias Schmidt: initial version.

Examples

si <- c(1, 1, 3, 7, 8, 3, 8, 5, 2, 1)
cs_max <- 14.8
cs_width <- 20.0
fq_ref <- 600.25 * 1e6
fq_width <- 12005
spectrum <- make_spectrum(si, cs_max, cs_width, fq_ref, fq_width)
spectrum2 <- make_spectrum(si, cs_max, cs_width, fq_ref, fq_width = 12010, force = TRUE)

Metabodecon Models

Description

Fit (fit_mdm()) or cross-validate (benchmark()) a binary classification model built on a set of NMR spectra. Both run the full deconvolute -> align -> snap -> featurize -> fit pipeline with sensible defaults and expose only the parameters a typical user tunes; the classification backend is chosen via model. Power users who need to swap individual pipeline stages can call the internal engines fit_mdm_internal() / benchmark_internal(), which take pluggable decon_fun / align_fun / snap_fun / feat_fun / fit_fun / predict_fun arguments.

fit_mdm() runs the pipeline once, or iterates over the cartesian product of npmax / maxShift / maxCombine when any is a vector and returns the row with the highest acc (ties broken by auc). benchmark() wraps fit_mdm() in outer k-fold cross-validation to estimate end-to-end performance on held-out spectra.

Usage

fit_mdm(
  x,
  y,
  model = c("ranger", "lasso"),
  npmax = -1L,
  maxShift = -1L,
  maxCombine = 10L,
  nworkers = 1L,
  seed = 1L,
  verbosity = 1L,
  ...
)

benchmark(
  x,
  y,
  model = c("ranger", "lasso"),
  npmax = -1L,
  maxShift = -1L,
  maxCombine = 10L,
  nworkers = 1L,
  seed = 1L,
  verbosity = 2L,
  k = 3L,
  ...
)

Arguments

Value

fit_mdm() returns an object of class mdm with elements model (trained backend model of the best grid row), ref (a list list(align, snap) for prediction-time replay), params (resolved pipeline parameters of the best row), the scalar performance of the best row (acc, auc, acc_se, auc_se), and mog (the augmented grid with per-row performance).

benchmark() returns a list with elements models (one fitted model per outer fold), predictions (per-spectrum out-of-fold predictions), performance (per-fold acc / auc) and overall (pooled acc / auc).

Examples

# Small, fast illustrative run. `deg` restricts deconvolution to a single
# parameter set and scalar npmax/maxShift/maxCombine give a one-row model
# grid (`mog`); benchmark() does a single 2-fold cross-validation round.
i <- c(1:3, 51:53)                       # 3 spectra per class
x <- sim2[i]
y <- attr(sim2, "group")[i]
deg <- expand.grid(nfit=3, smit=1, smws=3, delta=1.6)
m <- fit_mdm(x, y, npmax=10L, maxShift=1L, maxCombine=2L, deg=deg, verbosity=0)
bm <- benchmark(x, y, npmax=10L, maxShift=1L, maxCombine=2L, deg=deg, k=2L, verbosity=0)
# `model = "lasso"` selects L1-penalised logistic regression instead.

S3 methods for mdm objects

Description

WARNING: These methods are experimental and must not be used in production. Their API is very likely to change in non-backwards-compatible ways over the next few weeks.

S3 methods for objects of class mdm and summary.mdm.

predict.mdm() predicts probabilities, classes, link scores, or all three from an mdm object. When newdata is a spectra object, the spectra are deconvoluted, aligned and snapped to the references stored in the model before prediction. When newdata is a numeric matrix, it is used directly as the feature matrix.

print.mdm() prints a compact model summary.

coef.mdm() returns lasso coefficients (or ranger importance).

plot.mdm() plots the lasso path (or ranger importance bars).

summary.mdm() builds a compact summary list.

print.summary.mdm() prints formatted output for summary.mdm objects.

Usage

## S3 method for class 'mdm'
predict(
  object,
  newdata,
  type = c("all", "prob", "class", "link"),
  nworkers = 1,
  verbosity = 1,
  ...
)

## S3 method for class 'mdm'
print(x, ...)

## S3 method for class 'mdm'
coef(object, ...)

## S3 method for class 'mdm'
plot(x, ...)

## S3 method for class 'mdm'
summary(object, ...)

## S3 method for class 'summary.mdm'
print(x, ...)

Arguments

Value

• predict: numeric vector of probabilities, classes, and/or link scores.

• print: invisibly returns x.

• coef: coefficient object from glmnet (or ranger importance).

• plot: invisibly returns NULL.

• summary: object of class summary.mdm.

• print.summary.mdm: invisibly returns x.

Metabodecon Classes and Helpers

Description

Metabodecon represents NMR data using a small set of S3 classes connected by cumulative inheritance. A raw spectrum has class "spectrum". After deconvolute() it gains class "decon2", so its class vector becomes c("decon2", "spectrum"). After align() it gains class "align", with class vector c("align", "decon2", "spectrum"). The corresponding collection classes follow the same pattern.

Every deconvoluted or aligned object is still a spectrum in the base::inherits() sense, so S3 generic behavior for spectrum or spectra also works at every stage. Element order may vary between versions; always access fields by name, e.g. x$si or x[["cs"]]. Elements marked optional may be absent or NULL.

Usage

is_spectrum(x)
is_spectra(x)
as_spectra(x, ...)
as_decon2(x)
as_decons2(x)
get_names(x, default = "spectrum_\045d")

Arguments

Value

is_spectrum() and is_spectra() return TRUE or FALSE. The ⁠as_*()⁠ functions return an object of the requested class. get_names() returns a character vector.

Singlet classes

• spectrum: A single NMR spectrum. Class vector: "spectrum". Constructed by read_spectrum(), make_spectrum(), or simulate_spectrum(). Carries the fields under Always present (spectrum) below.

• decon2: A single deconvoluted NMR spectrum. Class vector: c("decon2", "spectrum"). Produced by deconvolute(). In addition to the spectrum fields, a decon2 carries the Added by deconvolute() fields below.

• align: A single deconvoluted NMR spectrum whose peak positions have been aligned across a collection. Class vector: c("align", "decon2", "spectrum"). Produced by align(). Carries everything a decon2 does, plus the Added by align() fields below.

Collection classes

For each singlet class there is a collection class that wraps a list of those singlets:

• spectra: List of spectrum. Class vector "spectra".

• decons2: List of decon2. Class vector c("decons2", "spectra").

• aligns: List of align. Class vector c("aligns", "decons2", "spectra").

Collections inherit from "spectra", so generic methods written for spectra also work on decons2 and aligns. Constructed by read_spectra() (returns spectra), deconvolute() when given a spectra (returns decons2), and align() (returns aligns). Concatenation follows the cumulative rule: the result class is the most-general, least-specific class among the inputs. Mixing an align with a plain decon2 yields decons2; mixing any plain spectrum in yields spectra.

Always present (spectrum)

1. cs: Vector of chemical shifts in ppm. Same length as si.

2. si: Vector of signal intensities (au). si[i] is the intensity at cs[i].

3. meta: Optional list of metadata, e.g. name (spectrum name), path (source path), type (experiment type), fq (signal frequencies in Hz), mfs (magnetic field strength), or simpar (true Lorentz-curve parameters for simulated spectra).

Added by deconvolute()

A decon2 object additionally has:

1. args: List of deconvolution parameters used (nfit, smit, smws, delta, sfr, igrs, npmax, use_rust, verbose).

2. sit: Data frame of signal intensities after transformations: sm (smoothed), sup (superposition of fitted Lorentz curves), and supal (superposition of aligned Lorentz curves, added by align()).

3. peak: Data frame of peak triplets with columns center, left, right: integer indices into cs.

4. lcpar: Data frame of Lorentz-curve parameters. Always carries x0 (center in ppm), A (amplitude), lambda (half-width) and pcide (integer column index into cs for x0). After clupa() also x0al / pcial (post-CluPA center and cs index). After snap_to_ref() also x0sn / pcisn (post-snap center and cs index, with NA for peaks snapped beyond maxCombine). A and lambda are preserved through every stage.

Added by align()

An align object has the same fields as decon2, but with the alignment slots populated:

• lcpar$x0al: Peak Centers after CluPA alignment in ppm

• lcpar$pcial: Peak Centers after CluPA alignment as cs indices

• lcpar$x0sn: Peak Centers after reference snapping in ppm (NA when snapped out)

• lcpar$pcisn: Peak Centers after reference snapping as cs indices (NA when snapped out)

• sit$supal: Signal Intensities of the superposition of aligned Lorentz curves

Predicates

is_spectrum() and is_spectra() test inheritance from the base metabodeconplus classes. Since decon2 and align inherit from spectrum, and decons2 and aligns inherit from spectra, they satisfy these checks. To test for a specific lifecycle stage, use base::inherits() directly, e.g. inherits(x, "decon2") or inherits(x, "aligns").

Converters

as_spectra() turns a path, spectrum, or list of spectrum objects into a spectra collection. as_decon2() and as_decons2() are identity converters that validate their input.

Naming helpers

get_names() returns collection names by checking each element's metadata, each element's direct name, the list names, and finally generated default names.

Author(s)

2024-2025 Tobias Schmidt: initial version.

Examples

s <- sim[[1]]
inherits(s, "spectrum")
is_spectrum(s)

d <- deconvolute(s, sfr = c(3.55, 3.35))
class(d) # c("decon2", "spectrum")
inherits(d, "spectrum") # TRUE

ds <- deconvolute(sim[1:3], sfr = c(3.55, 3.35))
class(ds)              # c("decons2", "spectra")
as_spectra(s)
get_names(list(s, myspec = s))

Return Path to File or Directory in metabodeconplus Package

Description

Recursively searches for files or directories within the 'metabodeconplus' package that match the given name.

Usage

metabodeconplus_file(name = "sim_01")

Arguments

Value

The file or directory path.

Author(s)

2024-2025 Tobias Schmidt: initial version.

Examples

# Unambiguous paths
metabodeconplus_file("urine_1")
metabodeconplus_file("urine_1.dx")
metabodeconplus_file("sim/sim_01")

# Ambiguous paths (i.e. multiple matches)
metabodeconplus_file("sim")
metabodeconplus_file("urine")

# Non-existing paths (i.e. a character vector of length zero gets returned)
metabodeconplus_file("asdfasdf")

Peak feature matrix

Description

Thin wrapper around si_mat() suitable as the feat_fun argument of fit_mdm(). Equivalent to si_mat(x, igrs=igrs); the snapping that used to live here has moved into align() (reference-snapping stage).

Usage

peak_mat(x, igrs = list(), peakPos = NULL, ...)

Arguments

Value

A numeric matrix with spectra in rows and chemical shifts as colnames.

Author(s)

2024-2026 Tobias Schmidt: initial version.

Examples

decons <- deconvolute(sim[1:3], sfr=c(3.55, 3.35), verbose=FALSE)
aligned <- align(decons, maxShift=50, maxCombine=20, verbose=FALSE)
X <- peak_mat(aligned)

Plot Spectra

Description

Plot a set of deconvoluted spectra.

Usage

plot_spectra(
  x,
  foc_rgn = NULL,
  what = NULL,
  sfy = 1e+06,
  cols = NULL,
  lty = NULL,
  names = NULL,
  xlab = "Chemical Shift [ppm]",
  ylab = paste("Signal Intensity [au] /", sfy),
  mar = c(4.1, 4.1, 1.1, 0.1),
  lgd = TRUE,
  main = NULL,
  xaxt = "s",
  yaxt = "s"
)

Arguments

Value

A plot of the deconvoluted spectra.

Author(s)

2024-2025 Tobias Schmidt: initial version.

See Also

plot_spectrum() for a much more sophisticated plotting routine suitable for plotting a single spectrum.

Examples

x <- deconvolute(sim[1:4], sfr = c(3.55, 3.35))
plot_spectra(x)

Plot Spectrum

Description

Plot a spectrum and zoom in on a specific region.

Usage

plot_spectrum(
  x,
  ...,
  obj = x,
  foc_frac = get_foc_frac(obj),
  foc_rgn = get_foc_rgn(obj, foc_frac),
  sub1 = TRUE,
  sub2 = FALSE,
  sub3 = width(foc_rgn) < width(obj$cs),
  mar = NULL,
  frame = FALSE,
  con_lines = TRUE
)

Arguments

Details

This function first initializes a new plotting canvas. After that it calls draw_spectrum() multiple times to draw the following sub figures onto the plotting canvas:

1. The signal intensities in the focus region

2. The second derivative in the focus region

3. The signal intensities over all datapoints

The argument lists for the individual calls to draw_spectrum() are determined at runtime and depend on the arguments passed to plot_spectrum() as well as the currently active graphics device. To customize the appearance of the individual sub plots, you can overwrite each value passed to draw_spectrum() by providing a corresponding named element in sub1, sub2 or sub3.

A sketch of the resulting figure is shown below.

 __________________________________________
|        ______________1_____________      |
|       | Sub1: Signal Intensity in  |     |
|       | Focus Region               |     |
|       |             /\             |     |
|       |            /  \            |     |
|       |           /    \  /\       |     |
|     11|          /      \/  \      |7    |
|       |     /\  /            \     |     |
|       |    /  \/              \    |     |
|       |   /                    \   |     |
|       |__/___________0__________\__|     |
|       | Sub2: Second Derivative    |     |
|     11| in Focus Region            |7    |
|       |____________________________|     |
|                      3                   |
|    __________________3_________________  |
|   |  Sub3: Signal Intensity over all   | |
|   |  Datapoints     ________________   | |
| 5 |                | Focus Rectangle|  |1|
|   |     /\         |       /\       |  | |
|   |    /  \        |      /  \/\    |  | |
|   |   /    \   /\  |   /\/      \   |  | |
|   |__/______\_/__\_|__/__________\__|__| |
|______________________5___________________|

Note that the figure created by plot_spectrum() can be part of a multi-figure configuration as created when setting mfrow or mfcol via graphics::par(). Example:

_______________________________________
| Plot Spectrum with   | Other Figure  |
| sub3 = TRUE          | Other Figure  |
|      ___________     |  ___________  |
|     | Sub Fig 1 |    | | x      x  | |
|     |___________|    | |      x    | |
|     |_Sub_Fig_2_|    | |      x    | |
|   _________________  | |   x     x | |
|  |    Sub Fig 3    | | |      x    | |
|  |_________________| | |___________| |
|______________________|_______________|
| Some other Figure    | Plot Spectrum |
|                      | sub3 = FALSE  |
|  _________________   |  ___________  |
| |     ___         |  | | Sub Fig 1 | |
| | ___/   \___     |  | |           | |
| |/           \____|  | |___________| |
| |                 |  | | Sub Fig 2 | |
| |_________________|  | |___________| |
|______________________|_______________|

Value

NULL. Called for side effect of plotting as sketched in 'Details'.

Author(s)

2024-2025 Tobias Schmidt: initial version.

Examples

## 1. Prepare a deconvoluted spectrum as input

spec <- sim[[1]]
decon <- deconvolute(sim[1], sfr = c(3.55, 3.35))

## 2.1. Plot the full (non-deconvoluted) spectrum
## 2.2. Remove connecting lines, and focus on a specific region specified in ppm
## 2.3. Show second derivative and focus on a specific region specified as fraction
## 2.4. Change color of focus rectangle and margins of sub figure 1
## 2.5. Hide xlab and show second derivative
## 2.6. Change the figure region for sub figure 1

plot_spectrum(spec, sub1 = FALSE)
plot_spectrum(decon, foc_rgn = c(3.49, 3.45), con_lines = FALSE)
plot_spectrum(decon, sub2 = TRUE, foc_frac = c(0.40, 0.30))
plot_spectrum(decon,
    sub1 = list(mar = c(3, 6, 3, 6), lt_axis = list(col = "violet")),
    foc_rect = list(border = "violet", col = transp("violet")),
    con_lines = list(col = "violet")
)
plot_spectrum(decon,
    sub2 = TRUE,
    sub3 = list(bt_text = list(text = "")),
    frame = TRUE,
    con_lines = FALSE
)
plot_spectrum(decon, sub1 = list(fig_rgn_npc = c(0,1,.3,1), mar = c(0,5,0,0)))

Read one or more spectra from Disk

Description

read_spectrum() reads a single spectrum from disk and returns it as spectrum object. read_spectra() can be used to read multiple spectra at once and returns a spectra object.

Usage

read_spectra(
  data_path = pkg_file("example_datasets/bruker/urine"),
  file_format = "bruker",
  expno = 10,
  procno = 10,
  raw = FALSE,
  silent = TRUE,
  force = FALSE
)

Arguments

Value

A spectrum object as described in metabodeconplus-classes.

Author(s)

2024-2025 Tobias Schmidt: initial version.

Examples

relpath <- "example_datasets/bruker/urine"
urine <- system.file(relpath, package = "metabodeconplus")
urine_1 <- file.path(urine, "urine_1")
urine_2 <- file.path(urine, "urine_2")
x1 <- read_spectrum(urine_1)
x2 <- read_spectrum(urine_2)
xx <- read_spectra(urine)
str(xx)
str(x1)
stopifnot(all.equal(x1, xx$urine_1))

Read one or more spectra from Disk

Description

read_spectrum() reads a single spectrum from disk and returns it as spectrum object. read_spectra() can be used to read multiple spectra at once and returns a spectra object.

Usage

read_spectrum(
  data_path = metabodeconplus_file("bruker/sim/sim_01"),
  file_format = "bruker",
  expno = 10,
  procno = 10,
  raw = FALSE,
  silent = TRUE,
  force = FALSE
)

Arguments

Value

A spectrum object as described in metabodeconplus-classes.

Author(s)

2024-2025 Tobias Schmidt: initial version.

Examples

relpath <- "example_datasets/bruker/urine"
urine <- system.file(relpath, package = "metabodeconplus")
urine_1 <- file.path(urine, "urine_1")
urine_2 <- file.path(urine, "urine_2")
x1 <- read_spectrum(urine_1)
x2 <- read_spectrum(urine_2)
xx <- read_spectra(urine)
str(xx)
str(x1)
stopifnot(all.equal(x1, xx$urine_1))

The SAP Dataset

Description

The SAP Dataset consists of a single 'Simple-As-Possible' (SAP) spectrum. The purpose of the SAP spectrum is to provide a straightforward example that can be used to test and understand the deconvolution algorithm in detail.

Usage

sap

Format

An object of class spectra of length 1.

Details

The first (and only) spectrum within the SAP dataset contains 128 datapoints ranging from -6.4 to 6.4 ppm with four peaks. A rough sketch of the spectrum is shown below:

-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~
|      SFR      |               w               |     SFR      |
|               |  x           www       p      |              |
|~-~-~-~-~-~-~-~|~-|-|-~-~-~-~-~|~-~-~-~-|-~-~-~-~-~-~-~-~-~-~-~
|               |  | |          |        |      |
6.4             |  | 2.24       0.047    -2.22  -3.2
                |  2.61
                3.2

Signal-integral matrix

Description

Builds a per-spectrum peak-area matrix. Each row is a spectrum, each column is a chemical-shift datapoint. For each peak, the column is picked from lcpar$pcisn (post-snap) when available, else lcpar$pcial (post-CluPA), else lcpar$pcide (post-decon). Peaks with pcisn = NA (snapped beyond maxCombine) are skipped. Collisions on the same column have their A * pi summed.

si_mat() is intentionally a dumb peak-list rasterizer: all alignment (continuous shift via CluPA) and reference snapping must have happened upstream — typically inside align(). To build a feature matrix where every spectrum shares the same column grid, run align(x, maxShift, maxCombine) first.

Usage

si_mat(x, drop_zero = FALSE, igrs = list(), peakPos = NULL, ...)

Arguments

Value

A numeric matrix with one row per spectrum and length(x[[1]]$cs) columns (the full cs grid). Column names are ppm values; row names are spectrum names.

Author(s)

2024-2026 Tobias Schmidt: initial version.

Examples

decons <- deconvolute(sim[1:2], sfr=c(3.55, 3.35), verbose=FALSE)
aligned <- align(decons, maxShift=50, maxCombine=20, verbose=FALSE)
X <- si_mat(aligned)

The Sim Dataset

Description

A simulated dataset generated from the Blood dataset.

Usage

sim

Format

A spectra object consisting of 16 spectrum objects, where each spectrum contains 2048 datapoints ranging from 3.60 to 3.29 ppm. For details about spectrum and spectra objects see metabodeconplus-classes.

The Sim2 Classification Dataset

Description

A simulated two-group classification dataset for demonstrating fit_mdm() and benchmark(). It contains 100 simulated 1D NMR spectra split evenly into groups A and B, where 3 out of 25 peaks per spectrum differ between the groups: in group A, two peaks are scaled by 1.25 and one peak by 1/1.25 (\approx 0.80). Group B is left unmodified. The first spectrum (sim2_001) is constructed without any global or per-peak ppm jitter so it can serve as a clean unshifted alignment reference.

Usage

sim2

Format

A spectra object consisting of 100 spectrum objects, where each spectrum contains 2048 datapoints ranging from 3.59 to 3.28 ppm. The per-spectrum group labels are attached as attr(sim2, "group"), a named factor with levels A and B. For details about spectrum and spectra objects see metabodeconplus-classes.

Each spectrum's meta$simpar carries the standard fields (x0, A, lambda, noise) plus five sim2-specific fields: base_x0 (the 25 reference peak positions, identical across spectra), dx0 (per-peak jitter in ppm), gx0 (scalar global ppm shift), diff_AB (integer indices into base_x0 of the peaks that differ between groups), and ab_factors (the multiplicative factors applied to those peaks in group A). They satisfy x0[k] = base_x0[k] + dx0[k] + gx0.

attr(sim2, "true_x0") is a numeric vector with the post-alignment ppm positions of the discriminating peaks (one per diff_AB index), useful for highlighting them on plots of aligned feature matrices.

Details

Peak parameters (positions, areas, half-widths and noise) were chosen to match the values recovered by deconvoluting the sim dataset, which itself is derived from the Blood reference dataset (see sim). Concretely:

• 25 base peaks per spectrum with positions drawn uniformly in ⁠[3.37, 3.52]⁠ ppm. The reference-grid step is ⁠0.00015 ppm/dp⁠.

• Per-peak jitter with standard deviation 4 datapoints (\approx 0.00060 ppm) plus a per-spectrum global ppm shift with standard deviation 8 datapoints (\approx 0.00120 ppm) to mimic chemical shift variation between samples.

• Base areas drawn from a log-normal distribution centered around 2500 (in ppm-area units) and varied per spectrum by ⁠+/-60%⁠.

• Base half-widths drawn uniformly in ⁠[0.0009, 0.0013]⁠ ppm and varied per spectrum by ⁠+/-10%⁠.

• Gaussian noise with standard deviation 2200.

• In group A, three base peaks (indices simpar$diff_AB) have their areas multiplied by c(1.25, 1.25, 1/1.25) (stored in simpar$ab_factors). Group B is left unmodified. Spectrum sim2_001 is generated with dx0 = 0 and gx0 = 0 to provide a clean unshifted alignment reference.

Simulate a 1D NMR Spectrum

Description

Simulates a 1D NMR spectrum based on the provided parameters.

Usage

simulate_spectrum(
  name = "sim_00",
  seed = sum(utf8ToInt(name)),
  ndp = 2048,
  npk = 10,
  csres = 0.00015,
  cs = seq(from = 3.6, length.out = ndp, by = -csres),
  pkr = quantile(cs, c(0.25, 0.75)),
  fqref = 600252806.95,
  x0 = sort(runif(npk, pkr[1], pkr[2])),
  A = runif(npk, 2.5, 20) * 1000,
  lambda = runif(npk, 0.9, 1.3)/1000,
  noise = rnorm(length(cs), sd = 1200)
)

Arguments

Value

A spectrum object as described in metabodeconplus-classes.

Author(s)

2024-2025 Tobias Schmidt: initial version.

Examples

simA <- simulate_spectrum("simA")
simA_copy <- simulate_spectrum("simA")
simB <- simulate_spectrum("simB")
simC <- simulate_spectrum("simC", npk = 20)
plot_spectrum(simC)
if (!identical(simA, simA_copy)) stop()
if ( identical(simA, simB     )) stop()

Temporary Session Directory

Description

Returns the path to metabodeconplus's temporary session directory. This directory equals subdirectory 'metabodeconplus' of R's temporary session directory base::tempdir() plus additional path normalization.

Usage

tmpdir(subdir = NULL, create = FALSE)

Arguments

Value

Returns the path to the temporary session directory.

Author(s)

2024-2025 Tobias Schmidt: initial version.

See Also

datadir_temp() datadir_persistent()

Examples

tmpdir()
tmpdir("simulate_spectra")

Make transparent

Description

Make a color transparent by adding an alpha channel.

Usage

transp(col = "violet", alpha = 0.08)

Arguments

Value

A character string representing the color with an alpha channel.

Author(s)

2024-2025 Tobias Schmidt: initial version.

Examples

transp("violet", 0.08)
transp("black", 0.5)

Print the Structure of a Directory Tree

Description

Prints the structure of a directory tree up to a specified maximum level of depth. It lists all files and directories under the specified path, displaying them in a tree-like structure.

Usage

tree(
  path,
  max.level = 2,
  max.entries = Inf,
  show.counts = FALSE,
  files.first = FALSE,
  level = 0,
  prefix = ""
)

tree_preview(
  path,
  max.level = 1,
  max.entries = 9,
  show.counts = TRUE,
  files.first = TRUE
)

Arguments

Value

NULL, called for its side effect of printing the directory structure.

Author(s)

2024-2025 Tobias Schmidt: initial version.

Examples

metabodeconplus_dir <- system.file(package = "metabodeconplus")
tree(metabodeconplus_dir, max.level = 1)

Calculate the Width of a Numeric Vector

Description

Calculates the width of a numeric vector by computing the difference between the maximum and minimum values in the vector.

Usage

width(x)

Arguments

Value

The width of the vector, calculated as the difference between its maximum and minimum values.

Author(s)

2024-2025 Tobias Schmidt: initial version.

Examples

vec <- c(1, 3, 5, 7, 9)
width(vec)