Contents

library("ggplot2")
library("grid")
library("dplyr")
library("cowplot")
library("IHWpaper")
theme_set(theme_cowplot())

Some general preliminary work, define factors, colours, which methods should we consider conservative, which anticonservative.

methods_pretty <- c("BH", "Clfdr", "Greedy Indep. Filt.", "IHW", "IHW naive", "FDRreg", "LSL GBH", "SBH", "TST GBH")
colors <- scales::hue_pal(h = c(0, 360) + 15, c = 100, l = 65, h.start = 0,direction = 1)(10)

conservative_methods <- c("BH", "Clfdr", "IHW", "FDRreg", "LSL GBH")
conservative_idx <- match(conservative_methods, methods_pretty)

anticonservative_methods <- c("Greedy Indep. Filt.", "IHW naive", "SBH", "TST GBH")
anticonservative_idx     <- match(anticonservative_methods, methods_pretty)

1 Panels a,b

1.1 Data for panels a,b

null_grb_file <- system.file("simulation_benchmarks/result_files",
                        "ihw_null_simulation_benchmark_grb.Rds", package = "IHWpaper")
null_e3_file <- system.file("simulation_benchmarks/result_files",
                        "ihw_null_simulation_benchmark_E3.Rds", package = "IHWpaper")
null_file <- system.file("simulation_benchmarks/result_files",
                        "ihw_null_simulation_benchmark.Rds", package = "IHWpaper")
null_df <- rbind(readRDS(null_grb_file),
                 readRDS(null_e3_file),
                 readRDS(null_file)) %>% 
                 filter(fdr_method != "IHW") %>% # just show IHW with E1-E3 
                 mutate(fdr_method = ifelse(fdr_method=="IHW E3", "IHW", fdr_method),  
                           fdr_method = sapply(strsplit(fdr_method," 20"), "[",1),
                           fdr_method = factor(fdr_method, levels= methods_pretty)) 

1.2 Panel a)

panel_a_df <-  filter(null_df, fdr_method %in% anticonservative_methods)
last_vals_a <- group_by(panel_a_df, fdr_method) %>% summarize(last_vals = max(FDR)) %>%
               mutate(last_vals = last_vals + c(0, 0,0.05, -0.03), 
                    label = fdr_method,
                    colour = colors[anticonservative_idx])
## `summarise()` ungrouping output (override with `.groups` argument)
panel_a <- ggplot(panel_a_df, aes(x=alpha, y=FDR, col=fdr_method)) +
                         geom_line(size=1.2) +
                         geom_abline(linetype="dashed") + 
                         xlab(expression(bold(paste("Nominal ",alpha)))) + 
                         scale_x_continuous(limits= c(0.01,0.1), breaks=seq(0.01,0.09,length=5)) +
                         ylim(0,0.9) +
                         theme(plot.margin = unit(c(3, 7.5, .2, .2), "lines"))+
                         scale_color_manual(values=colors[anticonservative_idx])+
                         theme(axis.title = element_text(face="bold") )


panel_a <- pretty_legend(panel_a, last_vals_a, 0.102)
panel_a

1.3 Panel b)

panel_b_df <- filter(null_df, fdr_method %in% conservative_methods)

last_vals_b <- group_by(panel_b_df, fdr_method) %>% summarize(last_vals = max(FDR)) %>% 
             mutate(last_vals = last_vals +  c(0.005,0.005,-0.005, 0 ,-0.005 ), 
                    label = fdr_method,
                    colour = colors[conservative_idx])
## `summarise()` ungrouping output (override with `.groups` argument)
panel_b <- ggplot(panel_b_df, aes(x=alpha, y=FDR, col=fdr_method)) +
                         geom_abline(linetype="dashed") + 
                         geom_line(size=1.2) +
                         xlab(expression(bold(paste("Nominal ",alpha)))) + 
                         theme(plot.margin = unit(c(3, 7.5, .2, .2), "lines"))+
                         scale_color_manual(values=colors[conservative_idx])+
                         theme(axis.title = element_text(face="bold") )


panel_b <- pretty_legend(panel_b, last_vals_b, 0.102 )
panel_b

1.4 Panels a, b)

panel_ab <- plot_grid(panel_a, panel_b, labels=c("a)","b)"), vjust=4.5) 
          

panel_ab <- ggdraw(panel_ab) +
          geom_rect(aes(xmin=0,xmax=1,ymin=0,ymax=0.95),
                            color="black",alpha=0.0) +
          draw_label("Nulls only", x = 1, y = 1,
            vjust = 1, hjust = 1, size = 15, fontface = 'bold') +
          theme(plot.margin=unit(c(.2,.2,.2,.2),"cm"))

panel_ab

#ggsave(panel_ab, "null_all.pdf",width=11,height=5.5)

2 Panels c, d)

2.1 Data for panels c,d)

effsize_grb_file <- system.file("simulation_benchmarks/result_files",
                        "ihw_du_ttest_inform_simulation_benchmark_grb.Rds", package = "IHWpaper")
effsize_e3_file <- system.file("simulation_benchmarks/result_files",
                        "ihw_du_ttest_inform_simulation_benchmark_E3.Rds", package = "IHWpaper")
effsize_file <- system.file("simulation_benchmarks/result_files",
                         "ihw_du_ttest_inform_simulation_benchmark.Rds", package = "IHWpaper")
effsize_df <- rbind(readRDS(effsize_grb_file),
                    readRDS(effsize_file),
                    readRDS(effsize_e3_file)) %>%
              filter(fdr_method != "IHW") %>% # just show IHW with E1-E3 
              mutate(fdr_method = ifelse(fdr_method=="IHW E3", "IHW", fdr_method),  
                           fdr_method = sapply(strsplit(fdr_method," 20"), "[",1),
                           fdr_method = factor(fdr_method, levels= methods_pretty)) 

2.2 Panel c)

panel_c_df <- filter(effsize_df, fdr_method %in% conservative_methods)

last_vals_c <- group_by(panel_c_df, fdr_method) %>% summarize(last_vals =  FDR[which.max(eff_size)]) %>%
               mutate(last_vals = last_vals + c(0,0.005,-0.005, 0 ,-0.01 ), 
                      label = fdr_method,
                      colour = colors[conservative_idx])
## `summarise()` ungrouping output (override with `.groups` argument)
panel_c <- ggplot(panel_c_df, aes(x=eff_size, y=FDR, col=fdr_method)) +
                         geom_hline(yintercept=0.1, linetype="dashed") + 
                         geom_line(size=1.2) +
                         xlab("Effect size") + 
                         theme(plot.margin = unit(c(3, 7.5, .2, .2), "lines"))+
                         scale_color_manual(values=colors[conservative_idx])+
                         theme(axis.title = element_text(face="bold") )


panel_c <- pretty_legend(panel_c, last_vals_c, 2.52 )
panel_c

2.3 Panel d)

panel_d_df <- filter(effsize_df, fdr_method %in% conservative_methods)

last_vals_d <- group_by(panel_d_df, fdr_method) %>% summarize(last_vals = power[which.max(eff_size)]) %>%
               mutate(last_vals = last_vals + c(0,-0.015,-0.035, 0.035 ,+0.005 ), 
                      label = fdr_method,
                      colour = colors[conservative_idx])
## `summarise()` ungrouping output (override with `.groups` argument)
panel_d <- ggplot(panel_c_df, aes(x=eff_size, y=power, col=fdr_method)) +
                         geom_line(size=1.2) +
                         xlab("Effect size") + 
                         ylab("Power")+
                         theme(plot.margin = unit(c(3, 7.5, .2, .2), "lines"))+
                         scale_color_manual(values=colors[conservative_idx])+
                         theme(axis.title = element_text(face="bold") )

panel_d <- pretty_legend(panel_d, last_vals_d, 2.52 )
panel_d