How to add adjusted pvalue to the corrplot matrix?

Question

I used following code to create correlation matrix and generate pvalue. How do I create adjusted pvalue for each correlation?

library(corrplot)
#Running correlation
M = cor(mtcars)

#identifying pvalues
testRes = cor.mtest(mtcars, conf.level = 0.95)

#creating pairwise p value
pval = as.data.frame(as.table(testRes$p))
head(pval)

output is:

    Var1    Var2    Freq
   <fct>    <fct>   <dbl>
1   mpg     mpg     0.000000e+00
2   cyl     mpg     6.112687e-10
3   disp    mpg     9.380327e-10
4   hp      mpg     1.787835e-07
5   drat    mpg     1.776240e-05
6   wt      mpg     1.293959e-10

How do I calculate adjusted pvalue here and add it in the colimn after pvalue?

Answer 1

Actually it's just cbind(<res>, p.adj=p.adjust(<res>$p, method='BH')).

First the correlation p-values solution without using the corrplot package.

pm <- outer(mtcars, mtcars, Vectorize(\(x, y) cor.test(x, y)$p.value))
res <- as.data.frame(as.table(pm)) |> setNames(c("Var1", "Var2", "p"))

Now, cbind adj. p-values; I use the Benjamini-Hochberg aka false discovery rate (FDR) method. Also common is the "bonferroni" method; see ?p.adjust for options.

res <- cbind(res, p.adj=p.adjust(res$p, method='BH'))  ## here you could use any `res`, e.g. your `pval` object and `pval&Freq`

head(res)
#   Var1 Var2            p        p.adj
# 1  mpg  mpg 0.000000e+00 0.000000e+00
# 2  cyl  mpg 6.112687e-10 3.892817e-09
# 3 disp  mpg 9.380327e-10 5.404855e-09
# 4   hp  mpg 1.787835e-07 6.555396e-07
# 5 drat  mpg 1.776240e-05 3.770615e-05
# 6   wt  mpg 1.293959e-10 9.209941e-10

If you like pipes, you can do:

res <- outer(mtcars, mtcars, Vectorize(\(x, y) cor.test(x, y)$p.value)) |>
  as.table() |> as.data.frame() |> setNames(c("Var1", "Var2", "p")) |>
  {\(.) cbind(., p.adj=p.adjust(.$p, method='BH'))}()

head(res)
#   Var1 Var2            p        p.adj
# 1  mpg  mpg 0.000000e+00 0.000000e+00
# 2  cyl  mpg 6.112687e-10 3.892817e-09
# 3 disp  mpg 9.380327e-10 5.404855e-09
# 4   hp  mpg 1.787835e-07 6.555396e-07
# 5 drat  mpg 1.776240e-05 3.770615e-05
# 6   wt  mpg 1.293959e-10 9.209941e-10

Answer 2

You can use bcdstats to generate adjusted pvalue.

library(corrplot)
library(bcdstats)

M = cor(mtcars)
adjp = adjust.corr(M, adjust="fdr")

head(as.data.frame(as.table(adjp$P)))

output:

    Var1    Var2    Freq
   <fct>    <fct>   <chr>
1   mpg     mpg 
2   cyl     mpg   <.0001
3   disp    mpg   <.0001
4   hp      mpg   <.0001
5   drat    mpg   <.0001
6   wt      mpg   <.0001