In neurodata/r-mgc: Multiscale Graph Correlation

require(ggplot2)
require(reshape2)
require(grid)
require(gridExtra)
require(mgc)
require(ICC)
require(I2C2)
require(cowplot)
require(lolR)
require(dplyr)
require(tidyr)
require(tidyverse)

g_legend<-function(a.gplot){
  tmp <- ggplot_gtable(ggplot_build(a.gplot))
  leg <- which(sapply(tmp$grobs, function(x) x$name) == "guide-box")
  legend <- tmp$grobs[[leg]]
  return(legend)
}
acols <- c("#FF0000", "#BBBBBB", "#BBBBBB", "#9997c9", "#778899", "#778899", "#387DC3")
linetypes <- c(1, 1, 2, 1, 1, 2, 1)
names(linetypes) <- names(acols) <- c("Discr", "ICC, best", "ICC, worst", "I2C2", "ANOVA, best",
                                      "ANOVA, worst", "MANOVA")

One Sample Results

alpha = 0.05
one.sample.results <- readRDS('data/sims/discr_sims_os.rds') %>%
  dplyr::mutate(outcome = as.numeric(p.value < alpha), n=as.numeric(n))

one.sample.powers <- one.sample.results %>%
  group_by(alg, n, sim.name) %>%
  summarize(power=mean(outcome))
sim.names <- unique(one.sample.powers$sim.name)
abbr <- c("(A)", "(B)", "(C)", "(D)", "(E)")

plots <- list()
for (i in 1:length(sim.names)) {
  plots[[i]] <- ggplot(subset(one.sample.powers, sim.name == sim.names[i]), 
                       aes(x=n, y=power, linetype=alg, group=alg, color=alg)) +
    geom_line(size=1.2) +
    scale_x_continuous() +
    theme_bw() +
    scale_color_manual(name="Statistic", values=acols) +
    scale_linetype_manual(name="Statistic", values=linetypes) +
    ggtitle(paste(abbr[i], sim.names[i])) +
    xlab("Number of Samples") +
    ylab("Statistical Power") +
    ylim(0, 1)
}

leg <- g_legend(plots[[1]] + theme(legend.position="bottom"))

plots <- lapply(1:length(plots), function(i) {
  if (i != 1) {
    return(plots[[i]] +
             xlab("") + ylab("") + theme(legend.position=NaN) +
      theme(axis.text.x=element_text(color="#FFFFFF"),
            axis.text.y=element_text(color="#FFFFFF"))
    )
  } else {
    return(plots[[i]] + theme(legend.position=NaN)) 
  }
})
grid.arrange(arrangeGrob(grobs=(plots), nrow=1), leg, top=textGrob("One Sample Results"), nrow=2, heights=c(0.9, 0.1))

Two Sample Results

alpha = 0.05
two.sample.results <- readRDS('data/sims/discr_sims_ts.rds') %>%
  mutate(outcome = as.numeric(p.value < alpha), n=as.numeric(n))

two.sample.powers <- two.sample.results %>%
  group_by(alg, n, sim.name) %>%
  summarize(power=mean(outcome))

sim.names <- unique(two.sample.powers$sim.name)
abbr <- c("(A)", "(B)", "(C)", "(D)", "(E)")

plots <- list()
for (i in 1:length(sim.names)) {
  plots[[i]] <- ggplot(subset(two.sample.powers, sim.name == sim.names[i]), 
                       aes(x=n, y=power, linetype=alg, group=alg, color=alg)) +
    geom_line(size=1.2) +
    scale_x_continuous() +
    theme_bw() +
    scale_color_manual(name="Statistic", values=acols) +
    scale_linetype_manual(name="Statistic", values=linetypes) +
    ggtitle(paste(abbr[i], sim.names[i])) +
    xlab("Number of Samples") +
    ylab("Statistical Power") +
    ylim(0, 1)
}

leg <- g_legend(plots[[1]] + theme(legend.position="bottom"))

plots <- lapply(1:length(plots), function(i) {
  if (i != 1) {
    return(plots[[i]] +
             xlab("") + ylab("") + theme(legend.position=NaN) +
      theme(axis.text.x=element_text(color="#FFFFFF"),
            axis.text.y=element_text(color="#FFFFFF"))
    )
  } else {
    return(plots[[i]] + theme(legend.position=NaN))
  }
})

grid.arrange(arrangeGrob(grobs=(plots), nrow=1), leg, top=textGrob("Two Sample Results"), nrow=2, heights=c(0.9, 0.1))

neurodata/r-mgc documentation built on March 12, 2021, 9:45 a.m.