Let’s work on cleaning some more real data from a project currently being conducted by Stephen and Victoria.

library(tidyverse)

In this study, participants read short stories, word-by-word,. The stories tested whether people attributed AGENTIVE or LOCATIVE meanings to the preposition by. We used past-tense passivised verbs to test this.

There are two formats for the stories:

  1. PERSON_A was with PERSON_ B, when PERSON_C appeared.

    1a. BP.A condition: PERSON_A was verbed by PERSON_C (agent)
    1b. BP.L condition: PERSON_A was verbed by PERSON_B (locative)

  2. PERSON_A was with OBJECT_B, when PERSON_C verbed.

    2a. BP.A OBJECT_B was verbed by PERSON_A (locative)
    2b. BP.L OBJECT_B was verbed by PERSON_C (agentive)

After each story, participants were asked:

Q. WHO VERBED PERSON_A / OBJECT_B?

1. PERSON_C        2. PERSON_B / PERSON_A

For each person, we need to calculate

  1. Whether their choice reflects locative or agentive interpretation
  2. Reading time for the words immediately after by

Load in the data from here

by_raw <- read_csv('https://www.stephenskalicky.com/r_data/nes_byphrase_raw.csv')
## Rows: 13527 Columns: 56
## ── Column specification ────────────────────────────────────────────
## Delimiter: ","
## chr (53): rt, button_pressed, PROLIFIC_PID, subject, sex, nes, att, trial_it...
## dbl  (3): condition, trial_index, age
## 
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.

1. create an object by_trim from by_raw which contains these columns:

by_trim <- by_raw %>%
  dplyr::select(subject, button_pressed, stim_id, story_condition, trial_item, button_cresp, comp_answer, button_order)

2. Inspect story_condition. Filter the data so that it only contains the two conditions we need. Name your new object by_trim01

summary(as.factor(by_trim$story_condition))
## BP.A BP.L    L PRAC    T NA's 
## 2940 2940 2940 1764 2940    3
by_trim01 <- by_trim %>%
  dplyr::filter(story_condition == 'BP.L' | story_condition == 'BP.A')

# str_detect()
# by_trim[grep("BP", by_trim$story_condition),]

3. Inspect trial_item. Filter the data so that it only contains the comprehension question asking who the agent is. Name your new object by_trim02.

summary(as.factor(by_trim01$trial_item))
##    agent_question location_question             story temporal_question 
##              1470              1470              1470              1470
by_trim02 <- by_trim01 %>%
  dplyr::filter(trial_item == 'agent_question')

# by_trim02 <- by_trim01[grep("agent_question", by_trim01$trial_item),]
summary(as.factor(by_trim02$button_cresp))
##  AL  LA 
## 742 728
summary(as.factor(by_trim02$button_pressed))
##   0   1 
## 757 713

4. Inspect button_cresp and button_pressed.

button_cresp you the order in which the two options appeared:
1. `LA` = LOCATIVE | AGENT
2. `AL` = AGENT | LOCATIVE

button_pressed tells you which button was pressed:

1. 0 = first option
2. 1 = second option

Create a new object by_trim03. Create a new column bp_choice which has a value of agentive or locative, depending on how the participant answered the agentive question.

by_trim03 <- by_trim02 %>%
  dplyr::mutate(bp_choice = case_when(button_cresp == 'AL' & button_pressed == 0 ~ 'agentive',
                                      button_cresp == 'LA' & button_pressed == 1 ~ 'agentive',
                                      button_cresp == 'AL' & button_pressed == 1 ~ 'locative',
                                      button_cresp == 'LA' & button_pressed == 0 ~ 'locative'))

5. Recreate the following ggplot from by_trim03:

summary(as.factor(by_trim03$stim_id))
## S01 S02 S03 S04 S05 S06 S07 S08 S09 S10 S11 S12 S13 S14 S15 S16 S17 S18 S19 S20 
##  73  75  74  72  73  75  74  72  73  75  74  72  73  75  74  72  73  75  74  72
# x = story_condition
# y = bp_choice
# data = by_trim03

ggplot(data = by_trim03, aes(x = story_condition, y = bp_choice, color = bp_choice)) +
  facet_wrap(. ~ stim_id) + 
  labs(x = '', y = '') + 
  #geom_point()
  #geom_point(position = 'jitter')
  geom_jitter(alpha = .5) +
  theme_bw()

6. Create a new object named by_rt from by_raw. Select the following columns and filter the data to only inlcude the two values of story_condition we care about.

by_rt <- by_raw %>%
  dplyr::select(subject, stim_id, story_condition, rt30, rt31, rt32) %>%
  dplyr::filter(story_condition == 'BP.L' | story_condition == 'BP.A')

7. Inspect the data. Find a way to remove the rows that do not have reading time data. Save your new object as by_rt01

by_rt01 <- by_rt %>%
  dplyr::filter(rt30 != '"')
1. in each story, the 30th word was `by` (rt30)
2. for stim id 1-12, `rt31` = `the`, `rt32` = `noun` (e.g., 'the woman')
3. for stim id 13-20, rt31 = NP

8. Create a new variable after_by which contains the RT for the NP immediately after by, with the following contraints. Name your new object by_rt02

1. in each story, the 30th word was `by` (rt30)
2. for stim id 1-12, `rt31` = `the`, `rt32` = `noun` (e.g., 'the woman')
3. for stim id 13-20, rt31 = NP

Your after_by variable should be set to the value of rt31 for stim_id 13:20 and to the mean of rt31 and rt32 for stim_id 1:12. Make sure you convert rt30, rt31, and rt32 to numeric. Don’t use mean().

Filter after_by to only contains values < 2000.

# there are smarter ways to do it - this is quite lazy
object_stories = c('S13', 'S14', 'S15', 'S16', 'S17', 'S18', 'S19', 'S20')

by_rt02 <- by_rt01 %>%
  mutate_at(vars(rt30, rt31, rt32), as.numeric) %>%
  filter(as.numeric(rt31) < 2000) %>%
  filter(as.numeric(rt32) < 2000) %>% 
  dplyr::mutate(after_by = if_else(stim_id %in% object_stories, as.numeric(rt31), ((as.numeric(rt31)+as.numeric(rt32))/2)))

9. Recreate the following plot:

ggplot(by_rt02, aes(x = story_condition, y = after_by, fill = story_condition)) +
  geom_violin(alpha = .5) +
  geom_jitter(alpha = .5, width = .1) +
  labs(x = '' , y = 'Reading times for NP after by') +
  theme_bw()

10. Interpret the results in light of these questions: