3_soc-xsit-plots.Rmd

---
title: "Soc-Xsit Plots"
author: "Kyle MacDonald"
output: html_document
---

```{r global options, include=T, echo=F}
rm(list=ls())
knitr::opts_chunk$set(fig.width=8, fig.height=5, fig.crop = F, echo = F, cache = T,
                      warning=F, cache=F, message=F, sanitize = T)
```

This script creates the plots for the paper: "Social cues modulate the representations underlying cross-situational learning"

```{r libraries, echo=F, warning=F}
source("useful.R"); library(pander); library(magrittr)
library(gridExtra); library(tidyr); library(langcog)
library(directlabels); library(forcats); library(cowplot)
```

```{r custom plotting theme, include = FALSE}
theme_soc_xsit <- theme_set(theme_cowplot())

theme_soc_xsit <- theme_update(
  axis.title.x = element_text(colour="black",size=16, margin=margin(15,0,0,0),face="plain"),
  axis.title.y = element_text(colour="black",size=16,margin=margin(0,15,0,0),face="plain"),
  legend.position="top",
  legend.text = element_text(size=12),
  strip.text = element_text(size=12),
  strip.background = element_blank(),
  panel.margin = unit(1, "lines")
  )
```

Read in data from all 4 Experiments:

* Experiment 1: Large scale experiment manipulating attention and memory demands
* Experiment 2: Replication of Experiment 1 with more ecologically valid stimulus set
* Experiment 3: Parametric manipulation of cue reliablity
* Experiment 4: Fixed inspection time for each exposure trial: two lengths -- short (6 sec) vs. long (9 sec)

```{r read data, warning=F}
df_expt1 <- read.csv("../data/3_final-processed/soc-xsit-expt1-finalData.csv")
df_expt2 <- read.csv("../data/3_final-processed/soc-xsit-expt2-finalData.csv")
df_expt3 <- read.csv("../data/3_final-processed/soc-xsit-expt3-finalData.csv")
df_expt4 <- read.csv("../data/3_final-processed/soc-xsit-expt4-finalData.csv")
```

## Experiment 1

```{r expt1 test trials}
# just RT filter
df_expo_expt1 <- df_expt1 %>%  
  filter(trial_category == "exposure", include_good_rt_exposure == "include")

df_test_expt1 <- df_expt1 %>%  
  filter(trial_category == "test", include_good_rt_test == "include",
         correct_exposure == T | condition == "No-Social")

# subject and trial level filter
df_test_expt1_filt <- df_expt1 %>% 
  filter(trial_category == "test") %>% 
  filter(include_good_rt_test == "include", include_good_rt_exposure == "include",
         include_expo == "include" | condition == "No-Social") %>%
  filter(correct_exposure == T | condition == "No-Social")

df_expo_expt1_analysis <- df_expt1 %>%
  filter(trial_category == "exposure") %>% 
  filter(include_good_rt_exposure == "include", 
         include_expo == "include" | condition == "No-Social") %>%
  filter(correct_exposure == T | condition == "No-Social")
```

```{r expt1 accuracy on exposure trials}
ms_expo_expt1 <-  df_expo_expt1 %>%
  filter(condition == "Social") %>%
  group_by(numPic, intervalNum, subid) %>%
  summarise(accuracy_exposure = mean(correct_exposure, na.rm = T)) %>% 
  group_by(numPic, intervalNum) %>%
  multi_boot_standard("accuracy_exposure")
```

```{r expt 1 accuracy on exposure trials plot, eval = F}
### Gaze following on exposure trials in the social condition
ggplot(data=ms_expo_expt1, aes(x=intervalNum, y=mean)) + 
  geom_pointrange(aes(ymin=ci_lower, 
                      ymax=ci_upper), size=0.6) +
  geom_smooth(method = "lm", se = F) +
  geom_hline(aes(yintercept=1/numPic), linetype = "dashed") +
  scale_x_continuous(limits=c(-.9,8), breaks=c(0, 1, 3, 7)) +
  scale_y_continuous(limits=c(0,1)) +
  scale_colour_manual(values=c("#1f78b4", "red")) +
  facet_grid( ~ numPic) + 
  xlab("Intervening Trials") + 
  ylab("Prop. Chose Gaze Target") +
  labs(colour = "Condition") +
  labs(linetype = "Trial Type") +
  guides(linetype=FALSE) 
```

### Inspection time on exposure trials and Accuracy at test

```{r expt1 summarize rt on exposure trials}
ms_expo_rt_expt1 <-  df_expo_expt1_analysis %>%
  mutate(rt = rt + 2000,
         rt_sec = rt / 1000) %>% 
  group_by(numPic, intervalNum, condition, subid) %>%
  summarise(mean_rt = mean(rt_sec)) %>% 
  group_by(numPic, intervalNum, condition) %>%
  multi_boot_standard("mean_rt")
```

```{r expt1 plot rt on exposure trials}
ms_expo_rt_expt1$condition <- revalue(ms_expo_rt_expt1$condition,
                                      c("No-Social" = "No-Gaze","Social" = "Gaze"))

ms_expo_rt_expt1$intervalNum <- as.factor(ms_expo_rt_expt1$intervalNum)
ms_expo_rt_expt1$numPic <- paste(ms_expo_rt_expt1$numPic, "-Referents", sep="")
y_axis_labs <- c("0.0", "2.0", "4.0", "6.0", "8.0")

exp1_rt_exposure_plot <- ggplot(data=ms_expo_rt_expt1, 
                                aes(x=intervalNum, y=mean, color = condition, 
                                    group = condition, shape = condition)) + 
  geom_smooth(method = "loess", se = F, span = 4, size = 1) +
  geom_pointrange(aes(ymin=ci_lower, ymax=ci_upper), 
                  fill = "white", stroke = 1.5, size= 1, fatten = 1.5,
                  position = position_dodge(width = .15)) +
  scale_y_continuous(limits=c(0,8), breaks = c(0,2,4,6,8), labels = y_axis_labs) +
  scale_colour_manual(values=c("#2c7fb8", "#e34a33"),
                      name = "Condition") +
  scale_shape_manual(values = c(21, 19),
                     name = "Condition") +
  facet_grid( ~ numPic) + 
  xlab(NULL) + 
  ylab("Inspection Time (sec)") +
  labs(colour = "Condition",linetype = "Trial Type") +
  guides(linetype=F) +
  panel_border()
```

```{r expt1 summarize accuracy at test}
ms_test_expt1 <- df_test_expt1  %>%
  group_by(condition, intervalNum, numPic, trialType, subid) %>%
  summarise(accuracy = mean(correct), 
            exclusionary_criteria = "Trial Level") %>% 
  group_by(condition, intervalNum, numPic, trialType) %>%
  multi_boot_standard("accuracy")

ms_test_filt_expt1 <- df_test_expt1_filt  %>%
  group_by(condition, intervalNum, numPic, trialType, subid) %>%
  summarise(accuracy = mean(correct), 
            exclusionary_criteria = "Subject and Trial level") %>%
  group_by(condition, intervalNum, numPic, trialType) %>%
  multi_boot_standard("accuracy")

ms_test_all_expt1 <- rbind(ms_test_expt1, ms_test_filt_expt1)
```

```{r expt1 acc test plot, echo=F}
ms_test_filt_expt1$condition <- revalue(ms_test_filt_expt1$condition,
                                   c("No-Social" = "No-Gaze","Social" = "Gaze"))
ms_test_filt_expt1 %<>%
  ungroup() %>% 
  mutate(intervalNum = as.factor(intervalNum),
         numPic_name = as.factor(paste(ms_test_expt1$numPic, "-Referents", sep="")))
```

```{r}
expt1.acc.test <- ggplot(data=ms_test_filt_expt1, 
                         aes(x=intervalNum, y=mean, 
                             colour = condition, 
                             lineType = trialType)) + 
  geom_smooth(aes(group=interaction(trialType, condition), linetype = trialType), 
              method = "loess", se = F, span = 10) +
  geom_pointrange(aes(ymin=ci_lower, ymax=ci_upper, shape = condition), 
                  fill = "white", stroke = 1.5, size= 1, fatten = 2,
                  position = position_dodge(width = .2)) +
  scale_shape_manual(values = c(21, 19), name = "Condition") +
  geom_hline(aes(yintercept=1/numPic), linetype = "dashed") +
  scale_y_continuous(limits=c(0,1.05), breaks = c(0, .25, .5, .75, 1.0)) +
  scale_colour_manual(values=c("#2c7fb8", "#e34a33"), name = "Condition") +
  facet_wrap(~ numPic_name, ncol = 4) + 
  xlab("Interval Between Exposure and Test") + 
  ylab("Prop. Correct") +
  labs(colour = "Condition") +
  labs(linetype = "Trial Type") +
  guides(linetype=F, shape = F, color = F) +
  panel_border()
```

```{r e1 final plot, fig.height=7}
plot_grid(exp1_rt_exposure_plot, expt1.acc.test, ncol=1, labels = c("A", "B"))
```

## Experiment 2

Tested a subset of the referent/interval conditions: 4-referent and 0-/3-interval in order to replicate findings from Experiment 1 with a more ecologically valid social cue.

```{r expt2 expo filters}
# just RT filter
df_expo_expt2 <- filter(df_expt2,
                        include_good_rt == "include",
                        condition_trial == "social",
                        trial_category == "exposure")

# RT, subject level and trial level filter 
df_expo_expt2_filt <- filter(df_expt2,
                             trial_category == "exposure",
                             condition_trial == "social" & mean_acc_exp > 0.25,
                             include_good_rt == "include")

# filter that gets both social/no-social trials
df_expo_expt2_analysis <- df_expt2 %>%
  mutate(rt = rt + 2000) %>% 
  filter(trial_category == "exposure",
         mean_acc_exp > 0.25,
         correct_exposure == T | condition_trial == "no-social",
         include_good_rt == "include")
```

```{r expt2 test filters}
# just RT filter
df_test_expt2 <- df_expt2 %>%
  filter(trial_category == "test",
         include_good_rt == "include")

# RT, subject level and trial level filter 
df_test_expt2_filt <- df_expt2 %>%
  filter(trial_category == "test",
         mean_acc_exp > 0.25 ,
         include_good_rt == "include",
         correct_exposure == T | condition_trial == "no-social")
```

### Inspection time on exposure trials and accuracy at test

```{r expt2 acc expo}
# unfiltered
ms_expo_expt2 <- df_expo_expt2 %>%
  group_by(intervalNum, subid) %>%
  summarise(accuracy_exposure = mean(correct)) %>% 
  group_by(intervalNum) %>% 
  multi_boot_standard("accuracy_exposure") %>% 
  mutate(filter = "Unfiltered")

# filtered
ms_expo_expt2_filt <- df_expo_expt2_filt %>%
  group_by(intervalNum, subid) %>%
  summarise(accuracy_exposure = mean(correct)) %>% 
  group_by(intervalNum) %>% 
  multi_boot_standard("accuracy_exposure") %>% 
  mutate(filter = "Filtered")

ms_expo_all_expt2 <- rbind(ms_expo_expt2, ms_expo_expt2_filt)
```

```{r expt2 rt expo}
ms_rt_expo_expt2 <- df_expo_expt2_analysis %>%
  group_by(condition_trial, intervalNum, subid) %>%
  mutate(rt_sec = rt / 1000) %>% 
  summarise(rt_exposure = mean(rt_sec)) %>% 
  group_by(condition_trial, intervalNum) %>%
  multi_boot_standard("rt_exposure")
```

```{r expt2 acc test aggregate}
# unfiltered
ms_test_expt2 <- df_test_expt2 %>%
  group_by(trialType, condition_trial, intervalNum, subid) %>%
  summarise(accuracy = mean(correct)) %>% 
  group_by(trialType, condition_trial, intervalNum) %>% 
  multi_boot_standard("accuracy") %>% 
  mutate(filter = "Unfiltered")

# filtered (subject level)
ms_test_expt2_filt <- df_test_expt2_filt %>%
  group_by(trialType, condition_trial, intervalNum, subid) %>%
  summarise(accuracy = mean(correct)) %>% 
  group_by(trialType, condition_trial, intervalNum) %>% 
  multi_boot_standard("accuracy") %>% 
  mutate(filter = "Filtered")

ms_test_all_expt2 <- rbind(ms_test_expt2, ms_test_expt2_filt)
```

```{r expt2 plots, echo=F}
## inspection time plot
ms_rt_expo_expt2$condition <- revalue(ms_rt_expo_expt2$condition_trial,
                                      c("no-social" = "No-Gaze", "social" = "Gaze"))

y_axis_labs_2 <- c("0.0", "2.0", "4.0", "6.0")

expt2.rt.expo <- ggplot(data=ms_rt_expo_expt2, 
                        aes(x=as.factor(intervalNum), 
                            y=mean, color = condition, 
                            group = condition_trial)) + 
  geom_line(size = 1) +
  geom_pointrange(aes(ymin=ci_lower, ymax=ci_upper, shape = condition_trial), 
                  fill = "white", stroke = 1.5, size= 1, fatten = 2) +
  scale_shape_manual(values = c(21, 19), name = "Condition") +
  scale_y_continuous(limits=c(0,7), breaks = c(0,2,4,6), labels = y_axis_labs_2) +
  scale_colour_manual(values=c("#2c7fb8", "#e34a33")) +
  guides(shape=F, color=F) +
  labs(x = "Intervening Trials",
       y = "Inspection Time (sec)") +
  theme(plot.margin = margin(t = 10, r = 20, b = 10, l = 20, unit = "pt")) +
  annotate("text", x = 2, y = 5.7, label = "No-Gaze", color = "#2c7fb8", size = 6) +
  annotate("text", x = 2, y = 3, label = "Gaze", color = "#e34a33", size = 6) +
  panel_border()

## accuracy plot
ms_test_expt2_filt$condition_trial <- revalue(ms_test_expt2_filt$condition_trial,
                                              c("no-social" = "No-Gaze", 
                                                "social" = "Gaze"))

expt2.acc.test.line <- ggplot(data=ms_test_expt2_filt, 
                              aes(x=as.factor(intervalNum), 
                                  y=mean, colour = condition_trial, shape = condition_trial)) + 
  geom_line(aes(group=interaction(trialType, condition_trial),
                linetype = trialType), size = 1, position = position_dodge(width = 0.1)) +
  geom_pointrange(aes(ymin=ci_lower, ymax=ci_upper), 
                  fill = "white", stroke = 2, size= 1, fatten = 2,
                  position = position_dodge(width = 0.1)) +
  geom_hline(aes(yintercept=1/4), linetype = "dashed") +
  scale_y_continuous(limits=c(0,1)) +
  scale_colour_manual(values=c("#2c7fb8", "#e34a33"), name = "Condition") +
  scale_shape_manual(values = c(21, 19), name = "Condition") +
  labs(x = "Intervening Trials",
       y = "Prop. Correct",
       linetype = "Trial Type") +
  guides(linetype=F) +
  theme(plot.margin = margin(t = 10, r = 20, b = 10, l = 20, unit = "pt"),
        legend.position=c(0.8,0.88)) +
  annotate("text", x = 2.35, y = 0.7, label = "Same", color = "black", size = 6) +
  annotate("text", x = 2.35, y = 0.4, label = "Switch", color = "black", size = 6) +
  panel_border()
```

```{r expt2 final plot, fig.height=5, fig.width = 8}
e2_main_plot <- plot_grid(expt2.rt.expo, expt2.acc.test.line, 
                          ncol=2, labels = c('A', 'B'))
e2_main_plot

ggsave(plot = e2_main_plot, "../paper/figs/expt2_new.png", device = "png", width = 9, height = 5)
```

## Experiment 3

Manipulated cue reliablity with 5 levels of reliability -- 0%, 25%, 50%, 75%, and 100% -- to test if learners would show graded changes in amount of information stored during learning.

```{r e3 reorder levels of reliability}
df_expt3$prop_cond_clean <- fct_relevel(df_expt3$prop_cond_clean, 
                                        c("0%", "25%", "50%", "75%", "100%"))
```

### Performance at test as a function of reliability condition

```{r expt3 summarizing}
# Accuracy on familiarization trials in test block
ms_test_test <- df_expt3 %>%
  filter(trial_category == "test", block == "test", include_good_rt == "include") %>% 
  group_by(prop_cond_clean, trialType, subid) %>%
  summarise(accuracy = mean(correct, na.rm = T)) %>% 
  group_by(prop_cond_clean, trialType) %>% 
  multi_boot_standard("accuracy", na.rm=T)

# Analyze subject reported reliablity
ms_test_subj_rel <- df_expt3 %>%
  filter(experiment == "replication") %>% 
  mutate(rel_subj = as.numeric(as.character(rel_subj))) %>% 
  group_by(prop_cond_clean, subid) %>%
  summarise(subjective_reliability = mean(rel_subj, na.rm=T)) %>% 
  group_by(prop_cond_clean) %>% 
  multi_boot_standard("subjective_reliability", na.rm = T)

# Analyze subject reported reliablity
mss_subj_rel <- df_expt3 %>% 
  filter(experiment == "replication", is.na(rel_subj) == F) %>% 
  select(subid, rel_subj, prop_cond_clean, total_exposure_correct, 
         same_accuracy, switch_accuracy, experiment) %>% 
  distinct()

# Reported reliability vs. number correct on exposure - means.
ms_test_subj_rel <- df_expt3 %>%
  filter(experiment == "replication", include_good_rt == "include", 
         is.na(rel_subj) == F) %>% 
  group_by(prop_cond_clean, total_exposure_correct, subid) %>%
  summarise(subj = mean(rel_subj, na.rm=TRUE)) %>% 
  group_by(prop_cond_clean, total_exposure_correct) %>% 
  multi_boot_standard("subj")

# Relationship between subjective reliability, reliablity condition, and test trial performance
gather_mss_rel_subj <- mss_subj_rel %>% 
  gather(key = trialType, value = sub_mean_accuracy, 5:6) %>% 
  arrange(subid)

gather_mss_rel_subj$rel_bin <- cut(gather_mss_rel_subj$rel_subj, breaks = 5, 
                                   labels = c("0-0.2", "0.2-0.4", "0.4-0.6", 
                                              "0.6-0.8", "0.8-1"))

ms_rel_subj <- gather_mss_rel_subj %>% 
  group_by(rel_bin, trialType, subid) %>% 
  summarise(mean_accuracy = mean(sub_mean_accuracy, na.rm = T)) %>% 
  group_by(rel_bin, trialType) %>% 
  multi_boot_standard("mean_accuracy", na.rm =T)

ms_acc_exp_test_expt3 <- df_expt3 %>%
  filter(trial_category == "test", block == "test",
         include_good_rt == "include", experiment == "replication") %>% 
  group_by(trialType, total_exposure_correct, subid) %>%
  summarise(correct = mean(correct)) %>% 
  group_by(trialType, total_exposure_correct) %>% 
  multi_boot_standard("correct")
```

```{r e4 inspection time and gaze following}
## correct as a fucntion of inspection time on exposure
df_inspect <- df_expt3 %>% 
  filter(trial_category == "exposure", block == "test", include_good_rt == "include") %>% 
  select(subid, itemNum, inspection_time_exposure = rt) %>% 
  mutate(inspection_time_exposure_sec = inspection_time_exposure / 1000) %>% 
  left_join(filter(df_expt3, trial_category == "test", block == "test"), 
            by = c("subid", "itemNum")) 

df_inspect %<>% mutate(exposure_bin = cut(inspection_time_exposure_sec, 5, 
                                          labels = c("0.0-2.5", "2.5-4.7", 
                                                     "4.7-6.9", "6.9-9.1", 
                                                     "9.1-11.4" )))

ms_inspect <- df_inspect %>% 
  group_by(trialType, exposure_bin, subid) %>% 
  summarise(mean_ss = mean(correct, na.rm = T)) %>% 
  group_by(trialType, exposure_bin) %>% 
  multi_boot_standard(column = "mean_ss", na.rm = T)

## Correct as a function of following gaze
ms_expt3_gf <- df_expt3 %>% 
  filter(trial_category == "test", include_good_rt == "include", block == "test") %>% 
  group_by(subid, trialType, correct_exposure, prop_cond_clean) %>% 
  summarise(mean_ss = mean(correct)) %>%
  group_by(trialType, correct_exposure, prop_cond_clean) %>%
  multi_boot_standard(column = "mean_ss")
```

```{r e3 main plots}
expt3_acc_test <- ggplot(data=ms_test_test, 
                         aes(x=prop_cond_clean, y=mean, group=trialType,
                             color=trialType, label = trialType)) + 
  geom_smooth(method='loess', se=F, span = 4) +
  geom_pointrange(aes(ymin=ci_lower, ymax=ci_upper, shape = trialType), 
                  fill = "white", stroke = 1.8, size= 1.3, fatten = 1.5) +
  scale_shape_manual(values = c(21, 19)) +
  geom_hline(aes(yintercept=1/4), linetype = "dashed") +
  scale_y_continuous(limits=c(0,1)) +
  scale_x_discrete(expand = c(0.2,0)) +
  #scale_colour_manual(values=c("#2c7fb8", "#e34a33")) +
  scale_color_grey(start = 0, end = 0.6) +
  labs(x="Reliability of Gaze", y = "Prop. Correct",
       color = "Trial Type") +
  guides(color=F,shape=F) +
  theme(plot.margin = margin(t = 10, r = 10, b = 10, l = 10, unit = "pt")) +
  annotate("text", x = 6, y = 0.8, label = "Same", color = "black", size = 6) +
  annotate("text", x = 6, y = 0.4, label = "Switch", color = "black", size = 6) +
  panel_border()

expt3_gf_plot <- ggplot(aes(x = prop_cond_clean, y = mean, color = correct_exposure, 
                            shape = correct_exposure), data = ms_expt3_gf) + 
  geom_smooth(aes(group=interaction(correct_exposure, trialType), 
                  linetype = trialType), method = "loess", se = F, span = 10,
              position = position_jitterdodge(jitter.width = 0.2, dodge.width = 0.1)) +
  geom_pointrange(aes(ymin=ci_lower, ymax=ci_upper), 
                  fill = "white", stroke = 1.8, size= 1, fatten = 1.5,
                  position = position_jitterdodge(jitter.width = 0.2, dodge.width = 0.1)) +
  scale_shape_manual(values = c(21, 19), 
                     name = "Followed Gaze\n(Exposure)", 
                     labels = c("No", "Yes")) +
  scale_colour_manual(values=c("#2c7fb8", "#e34a33"), 
                      labels = c("No", "Yes"), 
                       name = "Followed Gaze\n(Exposure)") +
  scale_x_discrete(expand = c(0.2,0)) +
  geom_hline(aes(yintercept=1/4), linetype = "dashed") +
  scale_y_continuous(limits=c(0,1)) +
  labs(y = "Prop. Correct",
       x = "Reliability of Gaze") + 
  guides(linetype=F, col = guide_legend(ncol = 2)) +
  theme(legend.justification=c(1,0), legend.position=c(1,0),
        plot.margin = margin(t = 10, r = 10, b = 10, l = 10, unit = "pt")) +
  annotate("text", x = 6, y = 0.8, label = "Same", color = "black", size = 6) +
  annotate("text", x = 6, y = 0.4, label = "Switch", color = "black", size = 6) +
  panel_border()
```

```{r}
e3_main_plot <- plot_grid(expt3_acc_test, expt3_gf_plot, ncol = 2, labels = c('A' ,'B'))
e3_main_plot

ggsave(plot = e3_main_plot, "../paper/figs/expt3_main_plot.png", device = "png", width = 9, height = 5)
```

### Performance at test as a function of gaze use and subjective reliability

```{r e3 sub plots}
expt3_acc_test_chose_gazetar <- ggplot(data=ms_acc_exp_test_expt3, 
                                       aes(x=total_exposure_correct, y=mean,
                                           group=trialType, color=trialType, 
                                           label = trialType)) + 
  geom_smooth(method='loess', se=F, span = 4) +
  geom_pointrange(aes(ymin=ci_lower, ymax=ci_upper, shape = trialType), 
                  fill = "white", stroke = 1.8, size= 1.3, fatten = 1.5,
                  position = position_dodge(width=0.1)) +
  scale_shape_manual(values = c(21, 19)) +
  geom_hline(aes(yintercept=1/4), linetype = "dashed") +
  scale_x_continuous(limits=c(-0.2,8.2), breaks=c(0:8)) +
  scale_y_continuous(limits=c(0,1)) +
  #scale_colour_manual(values=c("orange", "springgreen4")) +
  scale_color_grey(start = 0, end = 0.6) +
  labs(x = "Num. Trials Used Gaze", 
       y ="Prop. Correct",
       color = "Trial Type") +
  guides(color=F,shape=F) +
  annotate("text", x = 7.3, y = 0.75, label = "Same", size = 6) +
  annotate("text", x = 7.3, y = 0.17, label = "Switch", size = 6) +
  panel_border()

expt3_subj_rel_test_plot <- ggplot(data=ms_rel_subj, 
                                   aes(x=rel_bin, y=mean, group=trialType,
                                       color=trialType, label = trialType)) + 
  geom_smooth(method='loess', se=F, span = 4) +
  geom_pointrange(aes(ymin=ci_lower, ymax=ci_upper, shape = trialType), 
                  fill = "white", stroke = 1.8, size= 1.3, fatten = 1.5) +
  scale_shape_manual(values = c(21, 19)) +
  geom_hline(aes(yintercept=1/4), linetype = "dashed") +
  scale_y_continuous(limits=c(0,1)) +
  #scale_colour_manual(values=c("orange", "springgreen4")) +
  scale_color_grey(start = 0, end = 0.6) +
  labs(color = "Trial Type", 
       y = "Prop. Correct", 
       x = "Subjective Reliability") +
  guides(color=FALSE,shape=F) +
  annotate("text", x = 4.7, y = 0.75, label = "Same",  size = 6) +
  annotate("text", x = 4.7, y = 0.17, label = "Switch", size = 6) +
  panel_border()

expt3_inspect_plot <- ggplot(data=ms_inspect, 
                             aes(x=exposure_bin, y=mean, group=trialType,
                                 color=trialType, label = trialType)) + 
  geom_smooth(method='loess', se=F, span = 4) +
  geom_pointrange(aes(ymin=ci_lower, ymax=ci_upper, shape = trialType), 
                  fill = "white", stroke = 1.8, size= 1.3, fatten = 1.5) +
  scale_shape_manual(values = c(21, 19)) +
  geom_hline(aes(yintercept=1/4), linetype = "dashed") +
  scale_y_continuous(limits=c(0,1)) +
  scale_color_grey(start = 0, end = 0.6) +
  labs(x = "Inspection Time (sec)",
       y="Prop. Correct") +
  labs(color = "Trial Type") +
  guides(color=FALSE,shape=F) +
  annotate("text", x = 4.7, y = 0.75, label = "Same", size = 6) +
  annotate("text", x = 4.7, y = 0.35, label = "Switch", size = 6) + 
  panel_border()
```


```{r}
e3_sub_plots <- plot_grid(expt3_acc_test_chose_gazetar, expt3_subj_rel_test_plot, ncol = 2,
                          labels = c("A", "B"))
e3_sub_plots

ggsave(plot = e3_sub_plots, "../paper/figs/expt3_sub_plot.png", device = "png", width = 9, height = 5)
```

## Experiment 4

Fixed inspection time for exposure trials in the gaze and no-gaze conditions: two lengths -- short (6 sec) vs. long (9 sec). Tested for an effect of gaze over and above reduced inspection time.

```{r exp4 filter}
df_expt4_filtered <- df_expt4 %>% 
  filter(trial_category == "test", answer_type_exposure == "participant_response",
         correct_exposure == T | gaze_trial == "No-Gaze",
         mean_acc_exp > 0.25, include_good_rt == "include")
```

```{r e4 means 1, include = F}
# Get means and CIs for each condition (interval X trial type X inspection time)
ms_e4 <- df_expt4_filtered %>%
  group_by(subid, inspection_cond, intervalNum, trialType, gaze_trial) %>% 
  summarise(mean_correct = mean(correct)) %>% 
  group_by(inspection_cond, intervalNum, trialType, gaze_trial) %>% 
  multi_boot_standard(column = "mean_correct")

ms_e4$inspection_cond <- factor(ms_e4$inspection_cond, levels = c("short", "long"))
```

```{r e4 plot, include = F}
e4_plot <- ggplot(data=ms_e4, aes(x=as.factor(intervalNum), y=mean, 
                                  colour = gaze_trial, shape = gaze_trial)) + 
  geom_line(aes(group=interaction(trialType, gaze_trial), 
                linetype = trialType), size = 1, 
            position = position_dodge(width=.15)) +
  geom_pointrange(aes(ymin=ci_lower, ymax=ci_upper), 
                  fill = "white", stroke = 1.5, size= 1.3, fatten = 2.8,
                  position = position_dodge(width=.15)) +
  scale_shape_manual(values = c(19, 21)) +
  scale_colour_manual(values=c("#e34a33", "#2c7fb8")) +
  facet_wrap(~inspection_cond) +
  geom_hline(aes(yintercept=1/4), linetype = "dashed") +
  scale_y_continuous(limits=c(0,1)) +
  xlab("Intervening Trials") +
  ylab("Prop. Correct") +
  labs(colour = "Condition") +
  labs(linetype = "Trial Type") +
  guides(linetype=F, shape=F) +
  annotate("text", x = 2.35, y = 0.55, label = "Same", color = "black", size = 6) +
  annotate("text", x = 2.35, y = 0.3, label = "Switch", color = "black", size = 6)
```

```{r}
ggsave(plot = e4_plot, "../paper/figs/expt4.png", device = "png",
       width = 8, height = 5)
```

```{r e4 means}
ms_e4_2 <- df_expt4_filtered %>% 
  group_by(subid, intervalNum, trialType, gaze_trial) %>% 
  summarise(mean_correct = mean(correct)) %>% 
  group_by(intervalNum, trialType, gaze_trial) %>% 
  multi_boot_standard(column = "mean_correct")

ms_e4_2$gaze_trial <- factor(ms_e4_2$gaze_trial, levels = c("No-Gaze", "Gaze"))
```

### Performance at test as a function of interval and gaze condition

```{r e4 final plot, }
expt4.acc.test.line <- ggplot(data=ms_e4_2, 
                              aes(x=as.factor(intervalNum), y=mean, 
                                  colour = gaze_trial,shape = gaze_trial)) +
  geom_line(aes(group=interaction(trialType, gaze_trial), 
                linetype = trialType), size = 1) +
  geom_pointrange(aes(ymin=ci_lower, ymax=ci_upper), 
                  fill = "white", stroke = 1.5, size= 1, fatten = 2,
                  position = position_jitter(width=.02)) +
  scale_shape_manual(values = c(21, 19), name = "Condition") +
  scale_colour_manual(values=c("#2c7fb8", "#e34a33"), name = "Condition") +
  geom_hline(aes(yintercept=1/4), linetype = "dashed") +
  scale_y_continuous(limits=c(0,1)) +
  xlab("Interval Between Exposure and Test") +
  ylab("Prop. Correct") +
  guides(linetype=F) +
  annotate("text", x = 2.35, y = 0.6, label = "Same", color = "black", size = 6) +
  annotate("text", x = 2.35, y = 0.3, label = "Switch", color = "black", size = 6) +
  panel_border()
```

```{r, fig.width = 5}
expt4.acc.test.line
ggsave(plot = expt4.acc.test.line, "../paper/figs/expt4_collapsed.png", device = "png",
       width = 4.5, height = 5)
```