corx

Source: R/corx.R

Calculates correlations matrices. Relevant values are stored in a list with methods for easy retrieval and formatting in publication ready tables.

corx(
  data,
  x = NULL,
  y = NULL,
  z = NULL,
  method = c("pearson", "spearman", "kendall"),
  stars = c(0.05, 0.01, 0.001),
  p_adjust = c("none", "holm", "hochberg", "hommel", "bonferroni", "BH", "BY", "fdr"),
  round = 2,
  remove_lead = TRUE,
  triangle = NULL,
  caption = NULL,
  note = NULL,
  describe = FALSE,
  grey_nonsig = TRUE,
  call_only = FALSE
)

Arguments

data

data.frame or matrix

x

a vector of rownames. Defaults to all

y

a vector of colnames. If not supplied, y is set to x.

z

a vector of variable names. Control variables to be used in partial correlations - defaults to NULL

method

character. One of "pearson", "spearman", or "kendall"

stars

a numeric vector. This argument defines cut-offs for p-value stars.

p_adjust

character. What adjustment for multiple tests should be used? One of "none" (default), "holm", "hochberg", "hommel", "bonferroni", "BH", "BY", or "fdr"

round

numeric. Number of digits in printing

remove_lead

logical. if TRUE (the default), leading zeros are removed in summaries

triangle

character. one of "lower", "upper" or NULL (the default)

caption

character. table caption. Passed to plots

note

character. Text for a table note

describe

list of named functions. If functions are supplied to describe, new columns will be bound to the 'APA matrix' for each function in the list. Describe also accepts a variety of shortcuts. If describe is set to TRUE, mean and standard deviation are returned for all row variables. Describe can accept a character vector to call the following descriptive functions: c('mean','sd','var','median','iqr','skewness','kurtosis'). These shortcuts are powered by 'tidyselect'. Skewness and kurtosis are calculated using the 'moments' package. All functions retrieved with shortcuts remove missing values.

grey_nonsig

logical. Should non-significant values be grey in output? This argument does nothing if describe is not set to FALSE

call_only

logical. For debugging, if TRUE only the call is returned

Value

A list of class 'corx' which includes:

  • "call" The call which if evaluated reproduces the object

  • "apa" An 'APA' formatted correlation matrix with significance stars

  • "r" Raw correlation coefficients

  • "p" p-values

  • "n" Pairwise observations

  • "caption" Object caption

  • "note" Object note

Details

Constructs correlation matrices using 'stats::cor.test' unless z is specified. When z is specified ppcor::ppcor.test is used instead. Character and factor variables are not accepted. To prevent errors, users must first convert all variables to numeric.

Partial correlations

Supplying the argument z will call ppcor::pcor.test the correlation pair are supplied to arguments x and y. The vector of z given to corx is passed to argument z in pcor.test.

Missing data

Observations containing missing data required to complete a correlation or partial correlation are automatically removed.

P-adjust

P-values attained can be adjusted for multiple comparisons by using the 'p_adjust' argument. This calls the function stats::p.adjust. When a matrix is symmetrical, p-values are only adjusted for unique comparisons. When a correlation matrix is not symmetrical, all comparisons are assumed to be unique.

Examples

cor_mat <- corx(mtcars, x = c(mpg,cyl,disp), y = c(wt,drat,disp,qsec),
           z = wt, round = 2, stars = c(0.05),
           caption = "Controlling for weight" ,
           describe = list("mean" = function(x) mean(x,na.rm=TRUE)))
cor_mat
#> corx(data = mtcars, x = c(mpg, cyl, disp), y = c(wt, drat, disp, 
#>     qsec), z = wt, stars = c(0.05), round = 2, caption = "Controlling for weight", 
#>     describe = list(mean = function(x) mean(x, na.rm = TRUE)))
#> 
#> Controlling for weight
#> ---------------------------
#>      drat disp  qsec   mean
#> ---------------------------
#> mpg   .18 -.34  .55*  20.09
#> cyl  -.33 .72* -.74*   6.19
#> disp -.24   -  -.62* 230.72
#> ---------------------------
#> Note. * p < 0.05
coef(cor_mat)
#>            drat       disp       qsec
#> mpg   0.1806278 -0.3371636  0.5456251
#> cyl  -0.3260334  0.7235673 -0.7413854
#> disp -0.2404013  1.0000000 -0.6151686
cor_mat$p
#>            drat         disp         qsec
#> mpg  0.33085441 6.361981e-02 1.499883e-03
#> cyl  0.07346184 4.234471e-06 1.830433e-06
#> disp 0.19267873 1.000000e+00 2.306237e-04
plot(cor_mat)

cor_2 <- corx(iris[,-5], describe = c(median, IQR = iqr, kurt = kurtosis),
         note = "Using shortcuts to select describe functions", triangle = "lower")
cor_2
#> corx(data = iris[, -5], triangle = "lower", note = "Using shortcuts to select describe functions", 
#>     describe = c(median, IQR = iqr, kurt = kurtosis))
#> 
#> ------------------------------------------------------
#>                      1       2      3 median  IQR kurt
#> ------------------------------------------------------
#> 1. Sepal.Length     -                   5.80 1.30 2.43
#> 2. Sepal.Width    -.12      -           3.00 0.50 3.18
#> 3. Petal.Length .87*** -.43***     -    4.35 3.50 1.60
#> 4. Petal.Width  .82*** -.37*** .96***   1.30 1.50 1.66
#> ------------------------------------------------------
#> Note. Using shortcuts to select describe functions