pwcor, pwcov, pwnobs: (Pairwise, Weighted) Correlations, Covariances and Observation Counts

Description

Computes (pairwise, weighted) Pearsons correlations, covariances and observation counts. Pairwise correlations and covariances can be computed together with observation counts and p-values, and output as 3D array (default) or list of matrices. pwcor and pwcov offer an elaborate print method.

Notes: weights::wtd.cors is imported for weighted pairwise correlations (written in C for speed). For weighted correlations with bootstrap SE's see weights::wtd.cor (but bootstrap can be slow). Weighted correlations for complex surveys are implemented in jtools::svycor. An equivalent and faster implementation of pwcor (without weights) is provided in Hmisc::rcorr (written in Fortran)

Usage

pwcor(X, …, w = NULL, N = FALSE, P = FALSE, array = TRUE, use = "pairwise.complete.obs")
pwcov(X, …, w = NULL, N = FALSE, P = FALSE, array = TRUE, use = "pairwise.complete.obs")
pwnobs(X)
# S3 method for pwcor
print(x, digits = 2L, sig.level = 0.05, show = c("all","lower.tri","upper.tri"),
      spacing = 1L, …)
# S3 method for pwcov
print(x, digits = 2L, sig.level = 0.05, show = c("all","lower.tri","upper.tri"),
      spacing = 1L, …)

Arguments

a matrix or data.frame, for pwcor and pwcov all columns must be numeric. All functions are faster on matrices, so converting is advised for large data (see qM).

an object of class 'pwcor' / 'pwcov'.

numeric. A vector of (frequency) weights.

logical. TRUE also computes pairwise observation counts.

logical. TRUE also computes pairwise p-values (same as cor.test and Hmisc::rcorr).

array

logical. If N = TRUE or P = TRUE, TRUE (default) returns output as 3D array whereas FALSE returns a list of matrices.

use

argument passed to cor / cov. If use != "pairwise.complete.obs", sum(complete.cases(X)) is used for N, and p-values are computed accordingly.

digits

integer. The number of digits to round to in print.

sig.level

numeric. P-value threshold below which a '*' is displayed above significant coefficients if P = TRUE.

show

character. The part of the correlation / covariance matrix to display.

spacing

integer. Controls the spacing between different reported quantities in the printout of the matrix: 0 - compressed, 1 - single space, 2 - double space.

…

other arguments passed to cor or cov. Only sensible if P = FALSE.

Value

a numeric matrix, 3D array or list of matrices with the computed statistics. For pwcor and pwcov the object has a class 'pwcor' and 'pwcov', respectively.

Examples

Run this code

# NOT RUN {
mna <- na_insert(mtcars)
pwcor(mna)
pwcov(mna)
pwnobs(mna)
pwcor(mna, N = TRUE)
pwcor(mna, P = TRUE)
pwcor(mna, N = TRUE, P = TRUE)
aperm(pwcor(mna, N = TRUE, P = TRUE))
print(pwcor(mna, N = TRUE, P = TRUE), digits = 3, sig.level = 0.01, show = "lower.tri")
pwcor(mna, N = TRUE, P = TRUE, array = FALSE)
print(pwcor(mna, N = TRUE, P = TRUE, array = FALSE), show = "lower.tri")


# }

Run the code above in your browser using DataLab

Get 50% off unlimited learning