Generate data
Draw random uniform \(P\)-values
from [0, 0.05].
n_variants <- 10000
n_combine <- 3
pvalues <- tidyr::tibble(
id = rep(as.character(c(1:n_variants)), each = n_combine),
trait = rep(as.character(c(1:n_combine)), n_variants),
p = runif(n_variants * n_combine, min = 0, max = 1)
)
Combine \(P\)-Values with all
methods
Use the provided methods to compute the meta analysis \(P\)-values.
cauchy <- cauchy_combined(pvalues) %>%
rename(p_cauchy = p) %>%
select(-trait)
fisher <- fishers_combined(pvalues) %>%
rename(p_fisher = p) %>%
select(-trait)
harmonic <- harmonic_combined(pvalues) %>%
rename(p_harmonic = p) %>%
select(-trait)
min_max <- pvalues %>%
group_by(id) %>%
summarise(p_min = min(p), p_max = max(p))
combined_wide <- fisher %>%
left_join(harmonic) %>%
left_join(cauchy) %>%
left_join(min_max) %>%
select(-id)
# Joining with `by = join_by(id)`
# Joining with `by = join_by(id)`
# Joining with `by = join_by(id)`
Plot the result
The figure shows the data distribution on the diagonal and paired 2D
historgam plots for all combinations of \(P\)-values. The brighter yellows correspond
to higher counts, the green is in the middle of the scale, and the
darker blues correspond to the low values of the histogram.
my_bin <- function(data, mapping) {
ggplot(data = data, mapping = mapping) +
geom_bin2d() +
scale_fill_continuous(type = "viridis")
}
ggpairs(combined_wide, columns = 1:5,
lower = list(continuous = my_bin)) +
theme_bw() +
theme(axis.text.x = element_text(angle=90, hjust=1))
