Replace “Your Name” with your actual name.
In this lab assignment, you will practice interpreting interactions by visualizing them with different types of graphs. You’ll work with simulated datasets and explore interactions between categorical x categorical, linear x linear, and categorical x linear variables.
Please follow the instructions for each exercise, and use
ggplot2
for all visualizations.
Task: Use the dataset with two categorical variables and one outcome variable. Fit a model with a categorical x categorical interaction, and visualize the interaction using a bar graph with error bars.
Education_Level
(e.g., “High School”, “College”) and
Job_Type
(e.g., “Office”, “Field”), and an outcome variable
Job_Satisfaction
.Education_Level
and Job_Type
on
Job_Satisfaction
.emmeans
to compare all groups. First run
emmeans
and then pairs
.# Simulate data
set.seed(123)
Education_Level <- factor(rep(c("High School", "College"), each = 50))
Job_Type <- factor(rep(c("Office", "Field"), each = 25, times = 2))
Job_Satisfaction <- ifelse(Education_Level == "College",
70 + 10 * (Job_Type == "Office"),
60 + 5 * (Job_Type == "Field")) + rnorm(100, sd = 5)
data1 <- data.frame(Education_Level, Job_Type, Job_Satisfaction)
# Calculate means and standard errors with correct handling
means1 <- data1 %>%
group_by(Education_Level, Job_Type) %>%
summarise(
Job_Satisfaction_Mean = mean(Job_Satisfaction),
SE = (sd(Job_Satisfaction) / sqrt(n())), # Calculate SE
lower = Job_Satisfaction_Mean - 1.96 * SE, # Lower bound of the confidence interval
upper = Job_Satisfaction_Mean + 1.96 * SE # Upper bound of the confidence interval
)
# Check the calculated values
print(means1)
## # A tibble: 4 × 6
## # Groups: Education_Level [2]
## Education_Level Job_Type Job_Satisfaction_Mean SE lower upper
## <fct> <fct> <dbl> <dbl> <dbl> <dbl>
## 1 College Field 71.4 0.830 69.8 73.0
## 2 College Office 80.1 0.972 78.1 82.0
## 3 High School Field 65.5 0.919 63.7 67.3
## 4 High School Office 59.8 0.947 58.0 61.7
#Run your lm() model and summary() below using the data1 dataset
# Dummy Coding:
# High school is coded as 1, college is coded as 0
# Office is coded as 1, Field is coded as 0.
#Run your summary() below
# Fit linear model with interaction
model1 <- lm(Job_Satisfaction ~ Education_Level * Job_Type, data = data1)
summary(model1)
##
## Call:
## lm(formula = Job_Satisfaction ~ Education_Level * Job_Type, data = data1)
##
## Residuals:
## Min 1Q Median 3Q Max
## -11.5970 -2.8385 -0.2066 3.0467 10.3341
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 71.4129 0.9187 77.735 < 2e-16
## Education_LevelHigh School -5.9022 1.2992 -4.543 1.61e-05
## Job_TypeOffice 8.6383 1.2992 6.649 1.80e-09
## Education_LevelHigh School:Job_TypeOffice -14.3157 1.8373 -7.792 7.83e-12
##
## (Intercept) ***
## Education_LevelHigh School ***
## Job_TypeOffice ***
## Education_LevelHigh School:Job_TypeOffice ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 4.593 on 96 degrees of freedom
## Multiple R-squared: 0.7344, Adjusted R-squared: 0.7261
## F-statistic: 88.47 on 3 and 96 DF, p-value: < 2.2e-16
Interpret any significant Effects: College + Office workers report highest job satisfaction.
High School + Field workers have moderate satisfaction.
High School + Office workers do not benefit much from the office setting.
This suggests the benefits of Office jobs are more pronounced for College-educated workers — a classic interaction effect.
# Plot the bar graph of means with error bars.
# Plot the interaction with error bars
ggplot(means1, aes(x = Education_Level, y = Job_Satisfaction_Mean, fill = Job_Type)) +
geom_bar(stat = "identity", position = position_dodge(0.9), width = 0.7) +
geom_errorbar(aes(ymin = lower, ymax = upper),
width = 0.2, position = position_dodge(0.9)) +
labs(title = "Interaction Between Education Level and Job Type on Job Satisfaction",
x = "Education Level",
y = "Mean Job Satisfaction",
fill = "Job Type") +
theme_apa()
Interpretation of Plot:
Below run emmeans
to estimate the marginal means.
## Education_Level Job_Type emmean SE df lower.CL upper.CL
## College Field 71.4 0.919 96 69.6 73.2
## High School Field 65.5 0.919 96 63.7 67.3
## College Office 80.1 0.919 96 78.2 81.9
## High School Office 59.8 0.919 96 58.0 61.7
##
## Confidence level used: 0.95
Below run pairs
to perform pairwise comparisons (post
hoc tests) with Tukey adjustment for multiple comparisons.
## contrast estimate SE df t.ratio p.value
## College Field - High School Field 5.90 1.3 96 4.543 0.0001
## College Field - College Office -8.64 1.3 96 -6.649 <.0001
## College Field - High School Office 11.58 1.3 96 8.913 <.0001
## High School Field - College Office -14.54 1.3 96 -11.192 <.0001
## High School Field - High School Office 5.68 1.3 96 4.370 0.0002
## College Office - High School Office 20.22 1.3 96 15.562 <.0001
##
## P value adjustment: tukey method for comparing a family of 4 estimates
Task: Use the dataset with two continuous variables
and one outcome variable. Fit a model with a linear x linear
interaction, and visualize the interaction using a 2D plot with the
interactions
library.
Age
and Weekly_Hours_Worked
, and an outcome variable
Income
.Age
and Weekly_Hours_Worked
.interact_plot()
to visualize the
interaction.# Simulate data
set.seed(123)
Age <- rnorm(100, mean = 40, sd = 10)
Weekly_Hours_Worked <- rnorm(100, mean = 40, sd = 5)
Income <- 30000 + 500 * Age + 1000 * Weekly_Hours_Worked + 50 * Age * Weekly_Hours_Worked + rnorm(100, sd = 5000)
data2 <- data.frame(Age, Weekly_Hours_Worked, Income)
# Fit the lm() model
# summary() of the model
# Fit the lm() model
model2 <- lm(Income ~ Age * Weekly_Hours_Worked, data = data2)
summary(model2)
##
## Call:
## lm(formula = Income ~ Age * Weekly_Hours_Worked, data = data2)
##
## Residuals:
## Min 1Q Median 3Q Max
## -9360 -3389 -543 2948 11583
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 56644.49 18039.76 3.140 0.00225 **
## Age -182.83 446.51 -0.409 0.68311
## Weekly_Hours_Worked 397.91 461.04 0.863 0.39025
## Age:Weekly_Hours_Worked 65.91 11.45 5.757 1.03e-07 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 4734 on 96 degrees of freedom
## Multiple R-squared: 0.9678, Adjusted R-squared: 0.9668
## F-statistic: 962.5 on 3 and 96 DF, p-value: < 2.2e-16
# Load interactions library
library(interactions)
# 2D plot of linear x linear interaction
interact_plot(model2, pred = Weekly_Hours_Worked, modx = Age,
plot.points = TRUE, interval = TRUE)
Interpretation of the plot: The positive interaction between Age and Weekly_Hours_Worked suggests:
The effect of weekly work hours on income increases with age.
For older individuals, working more hours leads to higher gains in income compared to younger ones.
The plot should show steeper slopes for higher ages.
Task: Use the simulated dataset with one categorical variable and one continuous variable as predictors. Fit a model with a categorical x linear interaction, and visualize the interaction using an interaction plot.
Gender
and one continuous variable
Study_Hours
, and an outcome variable
Test_Score
.Gender
and Study_Hours
.ggplot2
to visualize
the interaction.# Simulate data
set.seed(123)
Gender <- factor(rep(c("Male", "Female"), each = 50))
Study_Hours <- rnorm(100, mean = 5, sd = 2)
Test_Score <- 60 + 10 * (Gender == "Female") + 5 * Study_Hours + 5 * (Gender == "Female") * Study_Hours + rnorm(100, sd = 5)
data3 <- data.frame(Gender, Study_Hours, Test_Score)
# Fit the model
#Dummy coding: Female coded as 0, Male coded as 1
# Summary of the model
# Fit model with Gender (Male = 1, Female = 0)
model3 <- lm(Test_Score ~ Gender * Study_Hours, data = data3)
summary(model3)
##
## Call:
## lm(formula = Test_Score ~ Gender * Study_Hours, data = data3)
##
## Residuals:
## Min 1Q Median 3Q Max
## -9.1191 -3.3752 -0.4846 3.0552 15.0753
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 72.3200 2.1265 34.009 < 2e-16 ***
## GenderMale -13.9836 2.9233 -4.784 6.22e-06 ***
## Study_Hours 9.5983 0.3805 25.223 < 2e-16 ***
## GenderMale:Study_Hours -4.5206 0.5323 -8.493 2.54e-13 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 4.824 on 96 degrees of freedom
## Multiple R-squared: 0.9624, Adjusted R-squared: 0.9613
## F-statistic: 820.2 on 3 and 96 DF, p-value: < 2.2e-16
Interpretation for significant main effects: Interpretation of Main Effects Study_Hours has a positive effect overall.
Gender effect: Females (coded as 0) have a higher baseline.
Interaction term: The effect of Study_Hours on Test_Score is stronger for females than for males.
# Plot the interaction
library(ggplot2)
ggplot(data3, aes(x = Study_Hours, y = Test_Score, color = Gender)) +
geom_point(alpha = 0.6) +
geom_smooth(method = "lm", se = TRUE) +
labs(title = "Interaction Between Gender and Study Hours on Test Score",
x = "Study Hours", y = "Test Score") +
theme_minimal()
## `geom_smooth()` using formula = 'y ~ x'
Interpretation of graph: For females, the slope is steeper: each extra hour of study yields a larger increase in test score.
Males benefit too, but less per hour.
Task: Given a multivariate dataset, create different types of graphs to visualize interactions and discuss which type of graph is most appropriate.
interact_plot
for the continuous x continuous interaction.
Use any variables you’d like as long as they fit the variable type.# Simulate data
set.seed(123)
Age <- rnorm(100, mean = 35, sd = 8)
Experience <- rnorm(100, mean = 10, sd = 5)
Gender <- factor(rep(c("Male", "Female"), each = 50))
Job_Type <- factor(rep(c("Office", "Field"), each = 25, times = 2))
Salary <- 30000 + 1000 * Age + 2000 * Experience + 150 * Age * Experience +
5000 * (Gender == "Female") + rnorm(100, sd = 5000)
data4 <- data.frame(Age, Experience, Gender, Job_Type, Salary)
head(data4)
## Age Experience Gender Job_Type Salary
## 1 30.51619 6.447967 Male Office 113921.3
## 2 33.15858 11.284419 Male Office 148415.8
## 3 47.46967 8.766541 Male Office 156098.7
## 4 35.56407 8.262287 Male Office 128880.7
## 5 36.03430 5.241907 Male Office 102779.7
## 6 48.72052 9.774861 Male Office 167324.5
## Age Experience Gender Job_Type Salary
## 95 45.88522 3.445992 Female Field 121507.64
## 96 30.19792 19.986067 Female Field 196034.22
## 97 52.49866 13.003544 Female Field 225240.32
## 98 47.26089 3.743643 Female Field 109532.84
## 99 33.11440 6.944170 Female Field 116600.46
## 100 26.78863 4.072600 Female Field 92548.31
# Categorical x Categorical Interaction (Bar graph with means and SE)
library(dplyr)
means4 <- data4 %>%
group_by(Gender, Job_Type) %>%
summarise(Salary_Mean = mean(Salary),
SE = sd(Salary)/sqrt(n()),
lower = Salary_Mean - 1.96 * SE,
upper = Salary_Mean + 1.96 * SE)
ggplot(means4, aes(x = Gender, y = Salary_Mean, fill = Job_Type)) +
geom_bar(stat = "identity", position = position_dodge(0.9)) +
geom_errorbar(aes(ymin = lower, ymax = upper),
position = position_dodge(0.9), width = 0.2) +
labs(title = "Categorical x Categorical Interaction: Gender × Job Type on Salary",
y = "Mean Salary") +
theme_minimal()
Graph interpretation: Females earn more across job types.
The difference between Job Types may differ by gender, suggesting an interaction.
# Categorical x Continuous plot
ggplot(data4, aes(x = Age, y = Salary, color = Gender)) +
geom_point(alpha = 0.5) +
geom_smooth(method = "lm", se = TRUE) +
labs(title = "Categorical × Continuous: Gender × Age on Salary",
y = "Salary", x = "Age") +
theme_minimal()
Graph interpretation: Salary increases with age, but the rate may be different between males and females.
Females may start higher or increase more steeply depending on the interaction term.
# Continuous x Continuous plot
model4 <- lm(Salary ~ Age * Experience, data = data4)
interact_plot(model4, pred = Age, modx = Experience,
plot.points = TRUE, interval = TRUE)
Graph interpretation: There is a synergistic effect: people with more experience and higher age earn significantly more.
Income rises more rapidly with age if experience is high.
Submission Instructions:
Ensure to knit your document to HTML format, checking that all content is correctly displayed before submission.