Safe correlation function which returns a sparse matrix without missing values

sparse_cor(
  x,
  y = NULL,
  method = "pearson",
  allow_neg = TRUE,
  remove_na = TRUE,
  remove_inf = TRUE,
  ...
)

Arguments

x: Sparse matrix or character vector.
y: Sparse matrix or character vector.
method: Method to use for calculating the correlation coefficient.
allow_neg: Logical. Whether to allow negative values or set them to 0.
remove_na: Logical. Whether to replace NA values with 0.
remove_inf: Logical. Whether to replace infinite values with 1.
...: Other arguments passed to cor function.

Value

A correlation matrix.

Examples

m1 <- simulate_sparse_matrix(
  1000, 1000,
  density = 0.01
)
m2 <- simulate_sparse_matrix(
  1000, 500,
  density = 0.01
)

all.equal(
  as.matrix(sparse_cor(m1)),
  cor(as_matrix(m1))
)
#> [1] TRUE
all.equal(
  as.matrix(sparse_cor(m1, m2)),
  cor(as_matrix(m1), as_matrix(m2))
)
#> [1] TRUE

system.time(
  sparse_cor(m1)
)
#>    user  system elapsed 
#>   0.041   0.013   0.053 
system.time(
  cor(as_matrix(m1))
)
#>    user  system elapsed 
#>   1.143   0.004   1.146 
system.time(
  sparse_cor(m1, m2)
)
#>    user  system elapsed 
#>   0.029   0.030   0.015 
system.time(
  cor(as_matrix(m1), as_matrix(m2))
)
#>    user  system elapsed 
#>   1.108   0.000   1.109 

# add missing values
m1[sample(1:500, 10)] <- NA
m2[sample(1:500, 10)] <- NA

sparse_cor(m1, m2)[1:5, 1:5]
#> 5 x 5 sparse Matrix of class "dgCMatrix"
#>        gene_1       gene_2       gene_3       gene_4       gene_5
#> gene_1      .  .            .            .            .          
#> gene_2      . -0.005809341 -0.005139148 -0.006487946 -0.003243621
#> gene_3      . -0.010749484 -0.009509372 -0.012005161 -0.006001929
#> gene_4      . -0.009950452 -0.008802521 -0.011112792 -0.005555794
#> gene_5      . -0.007806599 -0.006905993 -0.008718510 -0.004358782