Week 5 Homework

Homework Assignment for Week 5

For this homework assignment, we will continue to work with the dataset used in workshop 5 fake dataset.

Make sure you start a new RProject from scratch (call it homework_5 or something), as many of the steps will be similar but not identical to those we performed in workshop 5.

library(tidyverse)
library(here)
library(janitor)

d <- read_delim(here("Week 5 Homework/data", "workshop_5_data.csv")) |>
  clean_names()

Researchers noticed that the software they used to measure reaction time had a systematic error. Specifically, all reaction times that are larger than 800ms was actually 200ms larger than they should be. For example, if a particular trial was recorded as 1000ms, it actually should be corrected to 800ms.

1. Fix all of the reaction time trials so that these incorrect values are corrected, preserving the incorrect values in the dataset so that we can compare values if needed.

d_tidy <- d |>
  pivot_longer(cols = trial_1:trial_10,
               names_to = "trial_num",
               values_to = "rt")

d_tidy <- d_tidy |>
  mutate(rt_corrected = ifelse(rt > 800,
                               rt - 200,
                               rt))

2. How many trials needed to be fixed?

d_tidy <- d_tidy |>
  mutate(fixed = ifelse(rt_corrected == rt,
                        "no",
                        "yes"))

tally <- d_tidy |>
  group_by(fixed) |>
  summarize(n = n())

After correcting for this error, the researchers next want to exclude all reaction time values that are 1.5 standard deviations lower or higher than the global mean (i.e. the mean regardless of condition).

3. Remove all such outliers. How many trials of data were excluded because they were too low, and how many were excluded because they were too high?

bounds <- d_tidy |>
  group_by() |>
  summarize(mean_rt_corrected = mean(rt_corrected),
            sd_rt_corrected = sd(rt_corrected),
            min_rt = mean_rt_corrected - 1.5*sd_rt_corrected,
            max_rt = mean_rt_corrected + 1.5*sd_rt_corrected)

d_cleaned <- d_tidy |>
  mutate(outlier_type = ifelse(rt_corrected < bounds$min_rt,
                               "too_small",
                               "not_outlier"),
         outlier_type = ifelse(rt_corrected > bounds$max_rt,
                               "too_big",
                               outlier_type))

counts <- d_cleaned |>
  group_by(outlier_type) |>
  summarize(n = n())

4. Filter out the outliers, as well as those who did not provide a response for the enjoyment level question.

d_cleaned <- d_cleaned |>
  filter(outlier_type == "not_outlier",
         is.na(enjoyment_level) == FALSE)

Now, answer the following research questions, providing justification for how you drew your conclusions:

5. Do male participants have faster reaction times that female participants?

by_sex <- d_cleaned |>
  group_by(sex) |>
  summarize(mean_rt = mean(rt_corrected))

by_sex

## Male participants have slower reaction time on average

6. Did participants in the video games condition have faster reaction times than those in the TV condition?

by_condition <- d_cleaned |>
  group_by(condition) |>
  summarize(mean_rt = mean(rt_corrected))

by_condition

## TV condition had faster reaction time

7. Is there an interaction between sex and condition? If yes, describe the interaction.

interaction <- d_cleaned |>
  group_by(condition, sex) |>
  summarize(mean_rt = mean(rt_corrected))

interaction

## No, there is only a main effect of sex and condition

8. Consider the enjoyment level scale. Scores of 1-3 are considered to have “not enjoyed the activity”, 5-7 is considered “enjoyed the activity”, and 4 is considered “neutral”. Is there a relationship between enjoyment and participant’s reaction time?

# We start by recording

d_cleaned <- d_cleaned |>
  mutate(enjoyment_recoded = ifelse(as.numeric(enjoyment_level) < 4,
                                    "not_enjoyed",
                                    ifelse(as.numeric(enjoyment_level) > 4,
                                           "enjoyed",
                                           "neutral")))

relationship <- d_cleaned |>
  group_by(enjoyment_recoded) |>
  summarize(mean_rt = mean(rt_corrected))

relationship

## People who enjoyed the task had the slowest reaction time

## Bonus: We can plot this to visualize

d_cleaned$enjoyment_level <- as.numeric(d_cleaned$enjoyment_level)

plot_data <- d_cleaned |>
  group_by(subject_number, enjoyment_level) |>
  summarize(mean = mean(rt_corrected))

ggplot(data = plot_data,
       aes(x = enjoyment_level, y = mean)) +
  geom_point() +
  geom_smooth(method = "lm")