lsr (version 1.0.0)

correlate: Correlation matrices

Description

Computes a correlation matrix, optionally with hypothesis tests and corrections for multiple comparisons.

Usage

correlate(
  x,
  y = NULL,
  test = FALSE,
  corr.method = "pearson",
  p.adjust.method = "holm"
)

Value

Prints the correlation matrix. If test = TRUE, also prints a matrix of adjusted p-values and a matrix of sample sizes. The results are also returned as a list with four elements: correlation (the correlation matrix), p.value (the matrix of p-values), sample.size (the matrix of sample sizes), and args (a record of the options used). The list can be assigned to a variable and inspected if needed.

Arguments

x

A numeric vector, matrix, or data frame containing the variables to be correlated.

y

An optional second matrix or data frame. If provided, the variables in x are correlated with the variables in y rather than with each other.

test

Set to TRUE to display p-values and sample sizes alongside the correlations. Defaults to FALSE.

corr.method

The type of correlation to compute: "pearson" (the default), "spearman", or "kendall".

p.adjust.method

The method used to correct p-values for multiple comparisons. Defaults to "holm". All methods supported by p.adjust are accepted.

Details

Calculates a correlation matrix between all pairs of numeric variables. If only x is supplied, all pairwise correlations among the variables in x are computed. If both x and y are supplied, variables in x are correlated with variables in y.

Non-numeric variables (e.g., factors) are silently ignored: they appear in the output with NA in place of correlation values. This makes it convenient to pass an entire data frame without first removing categorical columns.

When test = TRUE, hypothesis tests are run for every pair of numeric variables. To reduce the risk of false positives from testing many pairs at once, p-values are adjusted using the Holm method by default. See p.adjust for other available methods.

Missing data are handled using pairwise complete cases, so sample sizes may differ across pairs of variables. If a particular pair of variables has too few complete observations for cor.test to run, the corresponding cell in the correlation matrix is left as NA rather than causing the whole call to fail.

See Also

cor, cor.test, p.adjust

Examples

# data frame with factors and missing values
data <- data.frame(
  anxiety = c(1.31, 2.72, 3.18, 4.21, 5.55, NA),
  stress = c(2.01, 3.45, 1.99, 3.25, 4.27, 6.80),
  depression = c(2.51, 1.77, 3.34, 5.83, 9.01, 7.74),
  happiness = c(4.02, 3.66, 5.23, 6.37, 7.83, 1.18),
  gender = factor(c("male", "female", "female", "male", "female", "female")),
  ssri = factor(c("no", "no", "no", NA, "yes", "yes"))
)

# Pearson correlation matrix (the default)
correlate(data)

# Spearman correlations
correlate(data, corr.method = "spearman")

# correlate two subsets of variables with each other
nervous <- data[, c("anxiety", "stress")]
happy <- data[, c("happiness", "depression")]
correlate(nervous, happy)

# include Holm-corrected p-values and sample sizes
correlate(data, test = TRUE)