Exploratory data analysis

Overview

Code 
library(tidyverse)
library(targets)
library(ggdist)
tar_source()
tar_load(c(exp1_data, exp2_data, exp1_data_agg, exp2_data_agg, exp3_data, exp3_data_agg))

data_agg <- bind_rows(
  mutate(exp1_data_agg, exp = "Exp 1"),
  mutate(exp2_data_agg, exp = "Exp 2")
)

Here I have a bunch of plots to help me understand the data for the later modeling.

Overall performance

Code 
data_agg |>
  mutate(gap = as.factor(gap)) |>
  ggplot(aes(gap, p_correct, color = chunk, group = chunk)) +
  geom_point() +
  geom_line() +
  scale_color_discrete("1st chunk LTM?") +
  facet_grid(exp ~ itemtype) +
  theme_pub()

Together to compare shared levels:

Code 
data_agg |>
  ggplot(aes(gap, p_correct, color = chunk, linetype = exp)) +
  geom_point() +
  geom_line() +
  scale_color_discrete("1st chunk LTM?") +
  scale_x_continuous(breaks = c(500, 3000, 6000)) +
  coord_cartesian(xlim = c(300, 6200)) +
  facet_wrap(~itemtype) +
  theme_pub()

Overall performance is a bit lower in experiment 2, but not by much. Still, it might be an issue if trying to model the two experiments together. The effect size of the chunk is also smaller.

Individual performance

Code 
data_agg_subj <- bind_rows(
  mutate(exp1_data, exp = "Exp 1"),
  mutate(exp2_data, exp = "Exp 2")
) |>
  group_by(id, exp) |>
  nest() |>
  mutate(data = map(data, aggregate_data)) |>
  unnest(data)

print overall accuracy for each subject:

Code 
data_agg_subj |>
  group_by(exp, id) |>
  summarize(p_correct = mean(p_correct)) |>
  ggplot(aes(exp, p_correct)) +
  geom_dotsinterval(side = "both") +
  theme_pub()

Experiment 3

Code 
exp3_data_agg |>
  ggplot(aes(gap, p_correct, color = chunk, group = chunk)) +
  geom_point() +
  stat_smooth(method = "lm", se = FALSE) +
  scale_color_discrete("1st chunk LTM?") +
  facet_wrap(~itemtype) +
  theme_pub()
`geom_smooth()` using formula = 'y ~ x'

Recall times

This is the distribution of total recall period duration for experiment 3:

Code 
total_rts <- exp3_data |>
  group_by(id, trial) |>
  summarize(total_recall_duration = sum(rt) + 9 * 0.2) |>
  pluck("total_recall_duration")
`summarise()` has grouped output by 'id'. You can override using the `.groups` argument.
Code 
hist(total_rts, breaks = 30, col = "grey", border = "white", xlab = "Total recall period duration (s.)", main = "")

Back to top