Create a functional data object

Creates an fdata object for 1D functional data (curves) or 2D functional data (surfaces). For 2D data, the internal storage uses a flattened matrix format [n, m1*m2] where each row represents a surface stored in row-major order.

Usage

fdata(
  mdata,
  argvals = NULL,
  rangeval = NULL,
  names = NULL,
  fdata2d = FALSE,
  id = NULL,
  metadata = NULL
)

Arguments

mdata

Input data. Can be:

• For 1D: A matrix [n, m] where n is number of curves, m is number of evaluation points

• For 2D: A 3D array [n, m1, m2] where n is number of surfaces, m1 x m2 is the grid size. Automatically detected and converted to flattened storage.

• For 2D: A matrix [n, m1*m2] (already flattened) with argvals specifying grid dimensions

• For 2D: A single surface matrix [m1, m2] with argvals specifying grid dimensions

argvals

Evaluation points. For 1D: a numeric vector. For 2D: a list with two numeric vectors specifying the s and t coordinates.

rangeval

Range of the argument values. For 1D: a numeric vector of length 2. For 2D: a list with two numeric vectors of length 2.

names

List with components 'main', 'xlab', 'ylab' for plot titles. For 2D, also 'zlab' for the surface value label.

fdata2d

Logical. If TRUE, create 2D functional data (surface). Automatically set to TRUE if mdata is a 3D array.

id

Optional character vector of identifiers for each observation. If NULL, uses row names of mdata or generates "obs_1", "obs_2", etc.

metadata

Optional data.frame with additional covariates (one row per observation). If metadata has an "id" column or non-default row names, they must match the id parameter.

Value

An object of class 'fdata' containing:

data

The data matrix. For 2D: flattened [n, m1*m2] format

argvals

Evaluation points

rangeval

Range of arguments

names

Plot labels

fdata2d

Logical indicating if 2D

dims

For 2D only: c(m1, m2) grid dimensions

id

Character vector of observation identifiers

metadata

Data frame of additional covariates (or NULL)

Details

For 2D functional data, surfaces are stored internally as a flattened matrix where each row is a surface in row-major order. To extract a single surface as a matrix, use subsetting with drop = TRUE or reshape manually:

# Extract surface i as matrix
surface_i <- fd[i, drop = TRUE]
# Or manually:
surface_i <- matrix(fd$data[i, ], nrow = fd$dims[1], ncol = fd$dims[2])

Examples

# Create 1D functional data (curves)
x <- matrix(rnorm(100), nrow = 10, ncol = 10)
fd <- fdata(x, argvals = seq(0, 1, length.out = 10))

# Create with identifiers and metadata
meta <- data.frame(group = rep(c("A", "B"), 5), endpoint = rnorm(10))
fd <- fdata(x, id = paste0("patient_", 1:10), metadata = meta)

# Access metadata
fd$id
#>  [1] "patient_1"  "patient_2"  "patient_3"  "patient_4"  "patient_5" 
#>  [6] "patient_6"  "patient_7"  "patient_8"  "patient_9"  "patient_10"
fd$metadata$group
#>  [1] "A" "B" "A" "B" "A" "B" "A" "B" "A" "B"

# Create 2D functional data from 3D array [n, m1, m2]
surfaces <- array(rnorm(500), dim = c(5, 10, 10))
fd2d <- fdata(surfaces)

# Access individual surface as matrix
surface_1 <- fd2d[1, drop = TRUE]