cor_matrix: Correlation matrix

In Deleetdk/kirkegaard: kirkegaard

Correlation matrix

Description

Outputs a correlation matrix. Supports weights, confidence intervals, correcting for measurement error and rounding.

Usage

cor_matrix(
  data,
  weights = NULL,
  by = NULL,
  reliabilities = NULL,
  CI = NULL,
  CI_template = "%r [%lower %upper]",
  skip_nonnumeric = T,
  CI_round = 2,
  p_val = F,
  p_template = "%r [%p]",
  p_round = 3,
  rank_order = F,
  asterisks = c(0.01, 0.005, 0.001),
  asterisks_only = T
)

Arguments

data	(data.frame or coercible into data.frame) The data.
weights	(numeric vector, numeric matrix/data.frame or character scalar) Weights to use for the correlations. Can be a numeric vector with weights, the name of a variable in data, or a matrix/data.frame with weights for each variable. If the latter, then harmonic means are used. If none given, defaults to rep(1, nrow(data)).
by	Grouping variable
reliabilities	(num vector) Reliabities used to correct for measurement error. If not present, assumed to be 1.
CI	(numeric scalar) The confidence level to use as a fraction.
CI_template	(character scalar) A template to use for formatting the confidence intervals.
skip_nonnumeric	(logical scalar) Whether to skip non-numeric variables. Defaults to TRUE.
CI_round	(whole number scalar) If confidence intervals are used, how many digits should be shown?
p_val	(log scalar) Add p values or not.
p_template	(chr scalar) If p values are desired, the template to use.
p_round	(int scalar) Number of digits to round p values to. Uses scientific notation for small numbers.
rank_order	(lgl or chr) Whether to use rank ordered data so as to compute Spearman's correlations instead.
asterisks	The thresholds to use for p value asterisks
asterisks_only	Whether to only include astrisks not numerical values

Details

Correction for measurement error is done using the standard Pearson formula: r_true = r_observed / sqrt(reliability_x * reliability_y).

Weighted correlations are calculated using wtd.cor or wtd.cors from weights package.

rank_order can take either a logical scalar or a character scalar. If given TRUE, it will use rank ranking method with the default settings (average ranks). If given a chr scalar, it will use that ranking method. If given FALSE, will not use rank data (default).

Confidence intervals are analytic confidence intervals based on the standard error.

Examples

cor_matrix(iris) #just correlations
cor_matrix(iris, CI = .95) #with confidence intervals
cor_matrix(iris, CI = .99) #with 99% confidence intervals
cor_matrix(iris, p_val = .95) #with p values
cor_matrix(iris, p_val = .95, p_template = "%r (%p)") #with p values, with an alternative template
cor_matrix(iris, reliabilities = c(.8, .9, .7, .75)) #correct for measurement error
cor_matrix(iris, reliabilities = c(.8, .9, .7, .75), CI = .95) #correct for measurement error + CI
cor_matrix(iris, rank_order = T) #rank order correlations, default method
cor_matrix(iris, rank_order = "first") #rank order correlations, specific method
cor_matrix(iris, weights = "Petal.Width") #weights from name
cor_matrix(iris, weights = 1:150) #weights from vector
#complex weights
cor_matrix(iris, weights = matrix(runif(nrow(iris) * 4), nrow = nrow(iris)))
cor_matrix(iris, weights = matrix(runif(nrow(iris) * 4), nrow = nrow(iris)), CI = .95)
#groups
cor_matrix(iris, by = iris$Species)
cor_matrix(iris, by = iris$Species, weights = 1:150)

Deleetdk/kirkegaard documentation built on April 1, 2024, 2:23 a.m.