NSCAW I In-Class Exercise

Author

Barboza-Salerno

Published

February 16, 2025

Introduction

This exercise explores child aggression and trauma using the NSCAW I dataset. Because this is designed for you to do, I supressed all of the output. The chunks of code give you a hints regarding the code that will answer each question. Create an R script and answer each question below. Cut and paste the chunk of code into your script, and fill it in so that it runs.

About the Data

The National Survey of Child and Adolescent Well-Being (NSCAW I) is a longitudinal study that examines children involved in the child welfare system. It collects child well-being indicators, caregiver characteristics, and service utilization data. The dataset includes information on trauma exposure, behavioral outcomes, and demographics.

For this assignment, we focus on three key variables:

Variable Type Description
tra1 Numeric Trauma: PTS T Score - This variable represents the Post-Traumatic Stress (PTS) T-score for the child. It measures trauma-related symptoms based on standardized assessments, with higher values indicating greater levels of trauma exposure or distress. Typically, scores above a certain threshold (e.g., 65-70) suggest clinically significant trauma symptoms.
bcagg1 Numeric Aggressive Behavior Standardized Score - This variable is a standardized measure of aggressive behavior in children, derived from behavioral assessments. Higher values suggest greater levels of externalizing behaviors such as verbal or physical aggression. Scores in the clinically significant range (e.g., above 65) may indicate the need for behavioral intervention.
ageY Numeric Child age in years

Instructions

Step 1: Load the Data

Make sure all of these libraries are installed or you will not be able to run the code. Also, make sure that your data set is in a subdirectory called data in the same directory as the one in which your R script is saved.

# Load necessary libraries
library(foreign)    # For reading SPSS files
library(dplyr)      # For data manipulation
library(ggplot2)    # For visualization
library(gtsummary)  # For creating regression tables
library(officer)    # For exporting tables to Word
library(flextable)  # For table formatting

# Load NSCAW I dataset
nscaw <- read.spss("data/nscaw.sav", to.data.frame = TRUE)

# Inspect the first few rows
head(nscaw)

Question: What type of variables are included in the dataset?

To get a sense of the variables in the dataset click here.

Step 2: Select Variables & Handle Missing Data

# Select relevant variables
nscaw_clean <- nscaw %>% 
  dplyr::select(FILL THE VARIABLES IN HERE) 

# Remove missing values
nscaw_clean <- na.omit(nscaw_clean)

# View summary statistics
summary(nscaw_clean)

Hint: Look at the summary statistics—do any variables have extreme values?

Step 3: Compute Correlations

# Compute correlation matrix
cor(nscaw_clean, use = "complete.obs")

Question: Which variables are most strongly correlated? Does the direction make sense?

Run the regression of child aggression on trauma score and age

Step 4: Run a Multiple Regression Model

# Run regression: Predict aggression using trauma score and age
reg_model <- lm(WRITE THE CODE TO RUN THE REGRESSION)

# Display model summary
summary(reg_model)

Hint: Examine the coefficients and p-values. Is trauma a significant predictor of aggression?

Step 5: Compute Fitted Values & Residuals

# Compute fitted values and residuals
nscaw_clean$fitted <- fitted(PROVIDE THE CODE HERE)
nscaw_clean$residuals <- residuals(reg_model)

# Compute standardized residuals
nscaw_clean$std_residuals <- PROVIDE THE CODE HERE(nscaw_clean$residuals)

Question: Why do we standardize residuals? What does this help us understand?

Step 6: Visualize Residuals

# Plot histogram of residuals
ggplot(nscaw_clean, aes(x = residuals)) +
  PROVIDE THE CODE HERE(bins = 30, fill = "lightblue", color = "black") +
  labs(title = "Histogram of Residuals", x = "Residuals", y = "Frequency")

Hint: Look for skewness—are residuals normally distributed?

# Plot standardized residuals vs. predicted values
ggplot(nscaw_clean, aes(x = fitted, y = std_residuals)) +
  geom_point(PROVIDE THE CODE HERE) +
  geom_hline(yintercept = 0, linetype = "dashed", color = "red") +
  labs(PROVIDE THE CODE HERE)

Hint: If you see patterns in the residuals, what assumption may be violated?

Step 7: Create & Export Regression Table

# Generate regression summary table
reg_table <- reg_model %>% 
  tbl_regression(
    label = list(tra1 = "PTS Score", bcagg1 = "Aggressive Behavior", ageY = "Age")
  ) %>% 
  modify_caption("Table 1: Regression Results for Aggression")

# Convert to flextable and export to Word
reg_flextable <- as_flex_table(reg_table)
doc <- read_docx()
doc <- body_add_flextable(doc, value = reg_flextable)
print(doc, target = "Regression_Results.docx")

Hint: Where does the p-value appear in the regression table? How does it inform significance?

Deliverables

Discuss the following:

  1. Correlation matrix output
  2. Regression model summary
  3. Histogram of residuals
  4. Residuals vs. fitted plot
  5. Regression results table (Word document)

Discussion Questions

  1. Based on your results, does trauma significantly predict aggression?
  2. How does child age influence the relationship between trauma and aggression?
  3. Are the residuals normally distributed? If not, what would you do differently?
  4. If the p-value for trauma is significant, what does that imply in real-world terms?
  5. How could you improve this model?

Final Thought: This exercise helps you develop a deeper understanding of how trauma influences child behavior. Future analyses could incorporate more variables (e.g., social support, environment) for better predictions.