Data Transformation - Practical

Process

flowchart TD
 A[Import data] --> B[Explore data]
 B --> C[Identify features for scaling]
 C--> D[Apply scaling]
 D --> E[Visualise and review]

  classDef level1 fill:#add8e6,stroke:#333,stroke-width:1px,font-weight:normal;
  classDef level2 fill:#fffacd,stroke:#333,stroke-width:1px,font-weight:normal;

  class A,B,C,D,E level1;
  
  

Data Import

Task

Download the dataset here:

https://www.dropbox.com/scl/fi/hrwuzi5kl3ftj4rmal5ew/scaling_practice.csv?rlkey=h94fsw9yp3mfw9yxlugq41qlw&dl=1

Solution

Code

# Load the dataset

sports_data <- read.csv("https://www.dropbox.com/scl/fi/hrwuzi5kl3ftj4rmal5ew/scaling_practice.csv?rlkey=h94fsw9yp3mfw9yxlugq41qlw&dl=1")

Data Exploration

Objective: To understand the structure and summary of the dataset.

Tasks

1. Inspect the structure of the dataset to identify variable types.
2. Summarise the dataset to understand the ranges and distributions of variables.

Solutions

Code

# Inspect structure
str(sports_data)

# Summary
summary(sports_data)

Identify Variables for Scaling

Objective: To determine which numeric variables are suitable for scaling.

Tasks

1. Identify numeric variables in the dataset that require scaling (e.g., using logical indexing).
2. Exclude non-numeric or categorical variables (such as Athlete_ID and Sport).
3. Create a subset of the dataset containing only the numeric variables to be scaled.

Solutions

Code

# Identify numeric variables
numeric_vars <- sports_data[, sapply(sports_data, is.numeric)]

# Exclude categorical variables
numeric_vars <- numeric_vars[, !names(numeric_vars) %in% c("Athlete_ID")]

# Display selected variables
names(numeric_vars)

Apply Scaling

Objective: To standardise numeric variables to have a mean of 0 and a standard deviation of 1.

Tasks

1. Scale the numeric variables using the scale() function.
2. Combine the scaled variables with the original dataset.
3. Ensure the combined dataset includes indicators for scaled variables.

Solutions

Code

# Scale the numeric variables
scaled_vars <- as.data.frame(scale(numeric_vars))

# Combine scaled variables with original dataset
sports_data_scaled <- cbind(sports_data, scaled_vars)

# Preview the combined dataset
head(sports_data_scaled)

Visualise and Review Scaled Data

Objective: To compare the original and scaled data.

Tasks

1. Generate summary statistics for the scaled variables.
2. Create boxplots to visualise distribution of original and scaled variables.
3. Compare and interpret the distributions.

Solutions

Code

library(reshape2) # needed for final figure
library(ggplot2)

# Summary statistics of scaled variables
summary(scaled_vars)

# Visualise original vs scaled data (example: Height_cm)
boxplot(sports_data$Height_cm, scaled_vars$Height_cm, 
        names = c("Original Height", "Scaled Height"), 
        main = "Comparison of Original and Scaled Height")

Code

# Visualise original and scaled data separately


# Melt the original and scaled datasets for plotting
original_melt <- melt(numeric_vars, variable.name = "Variable", value.name = "Original")
scaled_melt <- melt(scaled_vars, variable.name = "Variable", value.name = "Scaled")

# Plot original variables
original_plot <- ggplot(original_melt, aes(x = Variable, y = Original)) +
  geom_boxplot(fill = "blue", alpha = 0.5) +
  labs(title = "Original Variables", y = "Value", x = "Variable") +
  theme(axis.text.x = element_text(angle = 45, hjust = 1))

# Plot scaled variables
scaled_plot <- ggplot(scaled_melt, aes(x = Variable, y = Scaled)) +
  geom_boxplot(fill = "green", alpha = 0.5) +
  labs(title = "Scaled Variables", y = "Value", x = "Variable") +
  theme(axis.text.x = element_text(angle = 45, hjust = 1))

# Display the plots side by side
library(gridExtra)
grid.arrange(original_plot, scaled_plot, ncol = 2)