In LDSamson/associationstudies: Custom functions useful to perform association studies and prediction analyses

knitr::opts_chunk$set(echo = TRUE)
library(dplyr)
library(knitr)
library(kableExtra)
library(tidyr)

df <- data.frame(
  grouping_var = sample(c("batch1", "batch2", "batch3"), 100, TRUE),
  V1 = abs(rnorm(n = 100, mean = 10000, sd = 10000)),
  V2 = exp(rnorm(n = 100, mean = log(10^6), sd = log(10^4))),
  V3 = abs(rnorm(n = 100, mean = 100, sd = 10))
  )

sumtable <- data.frame(
  `V1 median (IQR)`   = format_values(median_iqr(df$V1), n.digits = 2),
  `V2 geomean (conf)` = format_values(gm_mean(df$V2), n.digits = 2),
  `V3 mean (SD)`      = format_values(mean_sd(df$V3), n.digits = 2),
  check.names = FALSE
)
sumtable

# create dummy data, with some variables having much higher average values
# than others, to illustrate the use of this function:
table_data <- immune_data %>%
  mutate(across(c(Neutrophils, Monocytes), .fns = ~(.+1)*10^5))

summ_table <- immune_data %>%
  group_by(Sex) %>%
  summarize(
    Neutrophils = format_values(gm_mean(Neutrophils)),
    Tregs = format_values(gm_mean(Tregs))
  )
knitr::kable(summ_table) %>%
  kableExtra::kable_styling()

# The function is also helpful to create many summary values at once:

table_data_long <- table_data %>%
  select(-c(Sex, Batch, Frailty.index)) %>%
  tidyr::pivot_longer(col = everything())

summ_table_long <- table_data_long %>%
  group_by(name) %>%
  summarize(
    "Summary value" = format_values(gm_mean(value, offset = 1), n.digits = 2)
  )
# offset gives better results when geomean and confidence interval are calculated
# and there are many zeros in the data. See details of the function gm_mean()
# for more information.

knitr::kable(summ_table_long) %>%
  kableExtra::kable_styling()

# dummy_data %>%
# #  group_by(grouping_var) %>%
#   summarize(format_values(gm_mean(V2)))
# gmmeanval <- gm_mean(dummy_data$V2)
# medianval <- median_iqr(dummy_data$V2)
#
# ggplot(dummy_data, aes(x = 1, y = V2)) +
#   geom_boxplot(outlier.size = NA, width = 0.1) +
#   geom_point(alpha = 0.2) +
#   geom_linerange(aes(x = 1, ymin = gmmeanval[2], ymax = gmmeanval[3]),
#                  col = "tomato", size = 2, position = position_nudge(x = 0.1)) +
#   # geom_linerange(aes(x = 1.05, ymin = medianval[1]-medianval[2]/2,
#   #                    ymax = medianval[1]+medianval[2]/4)) +
#   geom_point(aes(x = 1.05, y = medianval[1])) +
#   scale_y_log10()

LDSamson/associationstudies documentation built on April 14, 2025, 11:52 a.m.