In mbadge/hipsMultimodal: Analysis of MSH Hip Data

Objective

Characterize a cohort table.

library(knitr); library(pander); library(tibble); library(caret)
devtools::load_all()

knitr::opts_chunk$set(comment="#>",
                      fig.show='hold', fig.align="center", fig.height=8, fig.width=8,
                      message=FALSE,
                      warning=FALSE, cache=FALSE, tidy=TRUE, rownames.print=FALSE)

ggplot2::theme_set(vizR::theme_())

set.seed(123)

Cohorts are designated by a mutation and a filter, with all combinations valid. Mutations include binarization and imputation, which change the actual data values in the table. Filtering subsets the patients from a cross sectional study design to different case control study designs.

mutation: r params$mutating. (available: r lift_vd(str_x)("none", "binary", "complete"))
filter: r params$filtering. (available: r str_x(hipsOpt(cohorts)))

scalars <- hipsCohort_(params$mutating, params$filtering)

# After params$filtering
scalars[] %<>% map_if(.p = is.factor, .f = as.character)
scalars

Missing Data

# Filtered Scalars
DF.assess_missingness(scalars)
ggMissing(scalars)

Categorical Breakouts

# Identify categorical variables
catVars_chr <- scalars %>%
    keep(.p = ~class(.x) %in% c("factor", "character", "logical")) %>%
    keep(.p = ~n_distinct(.x) < 20) %>%
    names()

# Generate a summary table for the full dataset and each categorical grouping
walk(c(character(0), catVars_chr), ~print(Kable.cohort_df(x = scalars, grp = .x)))

Continuous Breakouts

DATA <- scalars
COL <- "sex"
ALPHA <- "fx"

scale_alpha_fx <- scale_alpha_manual("Fracture", values = c("TRUE"=1, "FALSE"=0.5))

# Identify continuous variables
contVars_chr <- DATA %>%
    keep(.p = is.numeric) %>%
    names()

# Melt table to tee up a facet_wrap by continuous variable
DATA %>%
    select(!!COL, !!ALPHA, one_of(contVars_chr)) %>%
    gather(key = "contVar", value = "contVal", one_of(contVars_chr)) %>%
    {
        ggplot(data=., aes_string(x = "contVal", alpha = ALPHA, fill = COL)) +
            geom_histogram(bins=15) +
            scale_alpha_fx +
            scale_fill_sex() +
            facet_wrap(~contVar, ncol=2, scales = "free")
    }