NB_avg %>%
  mutate(cluster = dbscan::dbscan(cbind(accuracy,rts), eps = 0.5, minPts = 3)$cluster) %>%
  ggplot(aes(targets, accuracy, color=factor(cluster))) +
  ggtitle("targets (window = 8 trials)", "NOTE: each point is a single participant") +
  geom_point(alpha=0.3) +
  #geom_smooth(method='lm', se = F) +
  facet_wrap(~condition)


NB_avg %>%
  ggplot(aes(lures, accuracy, color=condition)) +
  ggtitle("lures (window = 8 trials)", "NOTE: each point is a single participant") +
  geom_point(alpha=0.3) +
  geom_smooth(method='lm', se = F)

NB_avg %>%
  ggplot(aes(skewness, accuracy, color=condition)) +
  ggtitle("skewness (window = 8 trials)", "NOTE: each point is a single participant") +
  geom_point(alpha=0.3) +
  geom_smooth(method='lm', se = F)

NB_avg %>%
  ggplot(aes(lumpiness, accuracy, color=condition)) +
  ggtitle("lumpiness", "NOTE: each point is a single participant") +
  geom_point(alpha=0.3) +
  geom_smooth(method='lm', se = F)

NB_avg %>%
  ggplot(aes(lumpiness, rts, color=condition)) +
  ggtitle("lumpiness (window = 8 trials)", "NOTE: each point is a single participant") +
  xlab("lumpiness") +
  ylab("Average RT") +
  geom_point(alpha=0.3) +
  geom_smooth(method='lm', se = F)

nback <- NB_modified

nback %>%
  mutate(block=as.factor(block)) %>%
  mutate(trial=as.factor(trial)) %>%
  mutate(condition=ifelse(condition=='2-back',2,3)) %>%
  #filter(condition=='3-back') %>%
  #mutate(correct=as.numeric(correct)) %>%
  inspect_cor(show_plot = T)

averaged_nback <- NB_avg

averaged_nback %>%
  mutate(condition=ifelse(condition=='2-back',2,3)) %>%
  inspect_cor(show_plot = T)




base.df <- data.frame(x=100-base.roc$specificities,
                      y=base.roc$sensitivities,
                      auc = base.roc$auc[1],
                      model="base")

extd.df <- data.frame(x=100-extd.roc$specificities,
                      y=extd.roc$sensitivities,
                      auc = extd.roc$auc[1],
                      model="extended")

chance.df <- data.frame(x=1:100, y=1:100, model=" ", auc=50)

library(tidyverse)
library(ggrepel)
dats <- rbind(extd.df, base.df, chance.df)

to_auc_label <- function(model, auc) {
  paste(model,
        "\nAUC=",
        format(auc, digits=3),
        sep = ""
  )
}

dats$label = NA
dats[174,]$label = to_auc_label("Extended Model", dats[174,]$auc)
dats[647,]$label = to_auc_label("Base Model", dats[647,]$auc)


dats %>%
  ggplot(aes(x=x, y=y,
             group=model,
             color = model,
             linetype = factor(model))) +
  geom_path(size=0.8) +
  geom_label_repel(aes(label=label), na.rm = TRUE, box.padding = 1) +
  xlab("100% - Specificity") +
  ylab("Sensitivity") +
  theme_linedraw() +
  #scale_x_continuous(labels = scales::percent) +
  #scale_y_continuous(labels = scales::percent) +
  scale_fill_brewer(palette = "Greens") +
  scale_color_manual(values=c("black", "#808080", "gray")) +
  theme(legend.position = "none", text=element_text(size=16,family="Helvetica Neue Light"))

#library(extrafont)
#font_import(pattern = "Helvet.*")
ggsave("fig1.png", plot = last_plot(), width = 4, height = 4)
#embed_fonts("fig1.pdf")

library(latex2exp)

tex_labels <- c(
  TeX("Stimulus"),
  TeX("Vocabulary Size ($x_v$)"),
  TeX("Recent Lures Ratio"),
  TeX("Skewness"),
  TeX("Targets Ratio"),
  TeX("Lures Ratio"),
  TeX("$N$"),
  TeX("Recent Skewness"),
  TeX("Recent Lumpiness"),
  TeX("Recent Vocabulary Size"),
  TeX("Recent Targets Ratio")
)

feat_names <- c(
  "Stimulus",
  "Vocabulary Size",
  "Recent Lures Ratio",
  "Skewness",
  "Targets Ratio",
  "Lures Ratio",
  "N",
  "Recent Skewness",
  "Recent Lumpiness",
  "Recent Vocabulary Size",
  "Recent Targets Ratio"
)

#attStats(boruta_result) %>%
#  arrange(desc(meanImp)) %>%
boruta_scores %>%
  mutate(feature = rev(feat_names)) %>%
#  ggplot(aes(x=reorder(feature,meanImp,), y=meanImp, fill=decision, label=meanImp)) +
  ggplot(aes(x=reorder(feature,meanImp,), y=meanImp, label=meanImp)) +
  geom_bar(stat = "identity",width = 0.75) +
    #geom_text(size = 3, position = position_stack(vjust = 1.2)) + 
  ylab("Relative Importance Score") +
  xlab("Feature") +
  theme_linedraw() +
  scale_fill_manual(values=c("#303030","#a0a0a0")) +
  labs(fill = "Feature Selection Decision") +
  theme(
    text=element_text(size=16),
    axis.title.y =  element_blank(),
    axis.text.x = element_text(size = 14,colour = "#a0a0a0"),
    axis.text.y = element_text(size = 22),
    axis.title.x = element_text(size = 22),
    legend.position = "none",
#    legend.position = c(.95, .15),
    legend.justification = c("right", "top"),
    legend.box.just = "right",
    legend.margin = margin(5, 5, 5, 5)
  ) +
  coord_flip()


ggsave("fig2.png", plot = last_plot(), width = 9, height = 4.75)