---

Lewis and Frank (2018) raw data.

RAW_EXPERIMENTAL_DATA <- "no_dups_data_munged_A.csv"
EXPERIMENT_METADATA <- "experiment_key.csv" 

all_d <- read_csv(RAW_EXPERIMENTAL_DATA)

exp_key <- read_csv(EXPERIMENT_METADATA) %>% 
  mutate(order = gsub("\"", "", order), 
         exp = as.character(exp)) %>%
  mutate_if(is.character, as.factor) 

Munge data.

all_d_clean <- all_d %>%
  mutate(exp = as.character(exp),
         condition = fct_recode(condition, three_subordinate = "3sub",
                                three_basic = "3bas",
                                three_superordinate = "3sup")) %>%
  mutate_if(is.character, as_factor)  %>%
  select(exp, everything()) %>%
  left_join(exp_key %>% select(exp, order, timing))  %>%
  rename(stim_category = category,
         training_number = condition,
         presentation_style = timing)

kable(count(all_d_clean, training_number))

training_number	n
three_basic	1560
three_subordinate	1560
three_superordinate	1560
one	1560

Get subject mean across stimuli category.

ms <- all_d_clean %>%
  filter(training_number == "one" | training_number == "three_subordinate")  %>%
  gather(variable, value, c(prop_sub, prop_bas, prop_sup)) %>%
  filter(variable == "prop_bas")  %>%
  group_by(subids, presentation_style, order, training_number) %>% 
  summarize(prop_bas = mean(value))

Filter to “first” trials only.

group_first_ms <- ms %>%
 filter(
   (training_number == "three_subordinate" & order == "3-1") |
(training_number == "one" & order == "1-3")) %>%
  group_by(presentation_style, training_number) %>%
           tidyboot_mean(column = prop_bas, na.rm = TRUE)

ggplot(group_first_ms, 
       aes(x = presentation_style, y = mean, group = training_number, fill = training_number)) +
  geom_bar(position = "dodge", stat = "identity") +
  geom_linerange(aes(ymin = ci_lower, 
                     ymax = ci_upper), 
                 position = position_dodge(width = .9)) +
  ylim(0, 1) +
  ylab("Prop. basic-level choices") +
  theme_classic(base_size = 12)

kable(group_first_ms, digits = 2)

presentation_style	training_number	n	empirical_stat	ci_lower	mean	ci_upper
sequential	three_subordinate	122	0.30	0.24	0.30	0.36
sequential	one	129	0.66	0.60	0.66	0.73
simultaneous	three_subordinate	134	0.20	0.14	0.20	0.25
simultaneous	one	135	0.59	0.53	0.59	0.66

t-test of “three” trials

t.test(prop_bas ~ presentation_style, ms %>% 
          filter(training_number == "three_subordinate" & order == "3-1"))

## 
##  Welch Two Sample t-test
## 
## data:  prop_bas by presentation_style
## t = 2.5693, df = 243.99, p-value = 0.01079
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  0.02456515 0.18598048
## sample estimates:
##   mean in group sequential mean in group simultaneous 
##                  0.3005464                  0.1952736

t-test of “one” trials

t.test(prop_bas ~ presentation_style, ms %>% 
          filter(training_number == "one" & order == "1-3"))

## 
##  Welch Two Sample t-test
## 
## data:  prop_bas by presentation_style
## t = 1.4525, df = 260.72, p-value = 0.1476
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  -0.02494865  0.16523002
## sample estimates:
##   mean in group sequential mean in group simultaneous 
##                  0.6614987                  0.5913580