Correlation Plots

Download a copy of the vignette to follow along here: correlation_plots.Rmd

In this vignette, we go through how you can visualize associations between the variables included in your analyses.

Data set-up

library(metasnf)

# We'll just use the first few columns for this demo
abcd_cort_sa_minimal <- abcd_cort_sa[, 1:5]

# And one more mock categorical variable for demonstration purposes
abcd_city <- abcd_colour
abcd_city$"city" <- sample(
    c("toronto", "montreal", "vancouver"),
    size = nrow(abcd_city),
    replace = TRUE
)
abcd_city <- abcd_city |> dplyr::select(-"colour")

# Make sure to throw in all the data you're interested in visualizing for this
# data_list, including out-of-model measures and confounding variables.
data_list <- generate_data_list(
    list(abcd_cort_sa_minimal, "cortical_sa", "neuroimaging", "continuous"),
    list(abcd_h_income, "household_income", "demographics", "ordinal"),
    list(abcd_pubertal, "pubertal_status", "demographics", "continuous"),
    list(abcd_colour, "favourite_colour", "demographics", "categorical"),
    list(abcd_city, "city", "demographics", "categorical"),
    list(abcd_anxiety, "anxiety", "behaviour", "ordinal"),
    list(abcd_depress, "depressed", "behaviour", "ordinal"),
    uid = "patient"
)

summarize_dl(data_list)

##                              name        type       domain length width
## cortical_sa           cortical_sa  continuous neuroimaging     93     5
## household_income household_income     ordinal demographics     93     2
## pubertal_status   pubertal_status  continuous demographics     93     2
## favourite_colour favourite_colour categorical demographics     93     2
## city                         city categorical demographics     93     2
## anxiety                   anxiety     ordinal    behaviour     93     2
## depressed               depressed     ordinal    behaviour     93     2

# This matrix contains all the pairwise association p-values
assoc_pval_matrix <- calc_assoc_pval_matrix(data_list)

assoc_pval_matrix[1:3, 1:3]

##            mrisdp_303 mrisdp_304 mrisdp_305
## mrisdp_303  0.0000000  0.6374024  0.4513919
## mrisdp_304  0.6374024  0.0000000  0.2790341
## mrisdp_305  0.4513919  0.2790341  0.0000000

Heatmaps

Here’s what a basic heatmap looks like:

ap_heatmap <- assoc_pval_heatmap(
    assoc_pval_matrix
)

save_heatmap(
    ap_heatmap,
    "assoc_pval_heatmap.png",
    width = 650,
    height = 500,
    res = 100
)

Most of this data was generated randomly, but the “colour” variable is really just a categorical mapping of “cbcl_depress_r”.

You can draw attention to confounding variables and/or any out of model measures by specifying their names as shown below.

ap_heatmap2 <- assoc_pval_heatmap(
    assoc_pval_matrix,
    confounders = list(
        "Colour" = "colour",
        "Pubertal Status" = "pubertal_status"
    ),
    out_of_models = list(
        "City" = "city"
    )
)

save_heatmap(
    ap_heatmap2,
    "assoc_pval_heatmap2.png",
    width = 680,
    height = 500,
    res = 100
)

The ComplexHeatmap package offers functionality for splitting heatmaps into slices. One way to do the slices is by clustering the heatmap with k-means:

ap_heatmap3 <- assoc_pval_heatmap(
    assoc_pval_matrix,
    confounders = list(
        "Colour" = "colour",
        "Pubertal Status" = "pubertal_status"
    ),
    out_of_models = list(
        "City" = "city"
    ),
    row_km = 3,
    column_km = 3
)

save_heatmap(
    ap_heatmap3,
    "assoc_pval_heatmap3.png",
    width = 680,
    height = 500,
    res = 100
)

Another way to divide the heatmap is by variable domain. This can be done by providing a data_list with all the variables in the assoc_pval_matrix and setting split_by_domain to TRUE.

ap_heatmap4 <- assoc_pval_heatmap(
    assoc_pval_matrix,
    confounders = list(
        "Colour" = "colour",
        "Pubertal Status" = "pubertal_status"
    ),
    out_of_models = list(
        "City" = "city"
    ),
    data_list = data_list,
    split_by_domain = TRUE
)

save_heatmap(
    ap_heatmap4,
    "assoc_pval_heatmap4.png",
    width = 700,
    height = 500,
    res = 100
)