d2a {str2str}

Data-Frame to (3D+) Array or Matrix

Description

d2a converts a data.frame to a (3D+) array or matrix. This function assumes the data.frame contains 2+ variable dimensions, which will correspond to the returned arrays/matrix dimensions. One or multiple variables can contain the elements of the returned array (only one variable can contain the elements for returning a matrix). In the case of multiple variables, they will be binded as the last dimension in the returned array with dimnames equal to the variable names.

Usage

d2a(d, dim.nm = names(d)[-ncol(d)], rtn.dim.lab = "el_nm", check = TRUE)

Arguments

d

data.frame with at least 3 columns, where 2+ columns are variable dimensions and 1+ columns contain the to-be returned array/matrix elements.

dim.nm

character vector of 2+ length specifying the colnames in d that contain the variable dimensions. These do not need to be factors or character vectors. Note, all columns in d other than dim.nm are assumed to be element columns.

rtn.dim.lab

character vector of length 1 specifying the dimlabel to use for the last dimension in the returned array when there are multiple element columns in d. Note, that NA will be converted to "NA" and NULL will return an error. If you don't want any dimlabel to show, "" is probably the best option. If there is only one element column in d, this argument is ignored by d2a.

check

logical vector of length 1 specifying whether to check the structure of the input arguments. For example, check whether d is a data.frame. This argument is available to allow flexibility in whether the user values informative error messages (TRUE) vs. computational efficiency (FALSE).

Details

d2a is a wrapper for reshape::cast with the addition of reordering the dimnames by position, which sorts the dimnames by the position they first appear in the variable dimensions of the data.frame (reshape::cast sorts all the dimnames alphabetically).

Value

(3D+) array or matrix formed from the dimensions d[dim.nm] with dimlabels = dim.nm (and rtn.dim.lab if there are multiple element columns). The dimnames are the unique elements d[dim.nm] and are ordered by position (rather than alphabetically), which allow for conversions back to the original array after a call to a2d() or matrix after a call to m2d().

Examples

# 3D array
print(HairEyeColor)
d <- reshape::melt.array(HairEyeColor)
a <- reshape::cast(d, Hair ~ Eye ~ Sex)
identical(a, unclass(HairEyeColor)) # not the same as HairEyeColor
d <- a2d(HairEyeColor)
a <- d2a(d, dim.nm = c("Hair","Eye","Sex"))
identical(a, unclass(HairEyeColor)) # yes the same as HairEyeColor

# matrix
attitude_mat <- d2m(attitude)
d <- m2d(attitude_mat, col = 0)
m <- d2a(d)
identical(m, attitude_mat) # yes the same as attitude_mat

# correlation data.frame example for p-values using psych::corr.test(attitude[1:3])
# corr_test <- psych::corr.test(attitude)
# a <- lm2a(corr_test[c("r","se","t","p")])
r <- matrix(c(1.0000000, 0.8254176, 0.4261169, 0.8254176, 1.0000000, 0.5582882,
   0.4261169, 0.5582882, 1.0000000), nrow = 3, ncol = 3, byrow = FALSE)
se <- matrix(c(0.0000000, 0.1066848, 0.1709662, 0.1066848, 0.0000000, 0.1567886,
   0.1709662, 0.1567886, 0.0000000), nrow = 3, ncol = 3, byrow = FALSE)
t <- matrix(c(Inf, 7.736978, 2.492404, 7.736978, Inf, 3.560771,
   2.492404, 3.560771, Inf), nrow = 3, ncol = 3, byrow = FALSE)
p <- matrix(c(0.000000e+00, 1.987682e-08, 1.887702e-02, 5.963047e-08, 0.000000e+00,
   1.345519e-03, 0.018877022, 0.002691039, 0.000000000), nrow = 3, ncol = 3, byrow = FALSE)
a <- abind::abind(r, se, t, p, along = 3L)
dimnames(a) <- list(names(attitude[1:3]), names(attitude[1:3]), c("r","se","t","p"))
d <- a2d(a = a, col = 3)
a2 <- d2a(d = d, dim.nm = c("X1","X2"))
all.equal(a, a2) # dimlabels differ
dimnames(a2) <- unname(dimnames(a2))
all.equal(a, a2) # now it is true

# correlation data.frame example for confidence intervals using psych::corr.test(attitude[1:3])
# corr_test <- psych::corr.test(attitude[1:3])
# d <- corr_test[["ci"]][c("r","p","lower","upper")]
# cbind(d, after = 0L) <- reshape::colsplit(row.names(d), split = "-", names = c("X1","X2"))
# tmp <- d[c("X2","X1","r","p","lower","upper")]
# d2 <- plyr::rename(tmp, c("X1" = "X2", "X2" = "X1"))
# short_nm <- unique(c(fct2v(d[["X1"]]), fct2v(d[["X2"]])))
# d3 <- data.frame("X1" = short_nm, "X2" = short_nm,
#    "r" = NA_real_, "p" = NA_real_, "lower" = NA_real_, "upper" = NA_real_)
# d_all <- ld2d(ld = list(d, d2, d3), rtn.listnames.nm = NULL, rtn.rownames.nm = NULL)
d_all <- data.frame(
   "X1" = c("ratng","ratng","cmpln","cmpln","prvlg","prvlg","ratng","cmpln","prvlg"),
   "X2" = c("cmpln","prvlg","prvlg","ratng","ratng","cmpln","ratng","cmpln","prvlg"),
   "r" = c(0.8254176, 0.4261169, 0.5582882, 0.8254176, 0.4261169, 0.5582882, NA, NA, NA),
   "p" = c(1.987682e-08, 1.887702e-02, 1.345519e-03, 1.987682e-08,
      1.887702e-02, 1.345519e-03, NA, NA, NA),
   "lower" = c(0.66201277, 0.07778967, 0.24787510, 0.66201277, 0.07778967,
      0.24787510, NA, NA, NA),
   "upper" = c(0.9139139, 0.6817292, 0.7647418, 0.9139139, 0.6817292,
      0.7647418, NA, NA, NA)
)
tmp <- d2a(d = d_all, dim.nm = c("X1","X2"), rtn.dim.lab = "stat")
short_nm <- c("ratng","cmpln","prvlg")
dim_names <- list(short_nm, short_nm, c("r","p","lower","upper"))
a <- do.call(what = `[`, args = c(list(tmp), dim_names))
print(a)