Exercises with Solutions

Topics: exercises
Exercise 1: Data Visualization

Goal: Visualize the relationship between Age and Blood Pressure using a scatter plot.

Instructions:

  1. Use ggplot2.
  2. Add axis labels and a title.

Question:

What trend do you observe between Age and Blood Pressure across diagnoses?

library(ggplot2)

ggplot(df, aes(x = Age, y = Blood_Pressure, color = Diagnosis)) +
  geom_point(alpha = 0.7) +
  labs(
    title = "Age vs Blood Pressure by Diagnosis",
    x     = "Age",
    y     = "Blood Pressure (mmHg)"
  ) +
  theme_bw()
Exercise 2: Independent t-test

Goal: Compare if the average glucose of Diabetes and Healthy groups are significantly different.

Instructions:

  1. Extract glucose data for both groups.
  2. Use t.test().
  3. Print the t-statistic and p-value.

Question:

Is there a statistically significant difference at α = 0.05?

glucose_diabetes <- df$Glucose[df$Diagnosis == "Diabetes"]
glucose_healthy  <- df$Glucose[df$Diagnosis == "Healthy"]

result <- t.test(glucose_diabetes, glucose_healthy)

cat(sprintf("T-statistic: %.3f, P-value: %.4f\n",
            result$statistic, result$p.value))
Exercise 3: Logistic Regression

Goal: Predict whether a patient has a high glucose level (above median) using Age and Blood_Pressure.

Instructions:

  1. Create a binary column High_Glucose (1 = above median, 0 = below).
  2. Split the data into training/testing sets.
  3. Fit a logistic regression model and check accuracy.

Question:

How well does the model predict high-glucose patients based on age and blood pressure?

library(dplyr)

# Create target variable
median_glucose <- median(df$Glucose)
df <- df |> mutate(High_Glucose = as.integer(Glucose > median_glucose))

# Train/test split
set.seed(42)
train_idx  <- sample(seq_len(nrow(df)), size = floor(0.7 * nrow(df)))
train_data <- df[train_idx, ]
test_data  <- df[-train_idx, ]

# Train model
model <- glm(High_Glucose ~ Age + Blood_Pressure, data = train_data, family = binomial)

# Predict and evaluate
pred_prob  <- predict(model, newdata = test_data, type = "response")
pred_class <- ifelse(pred_prob > 0.5, 1, 0)
accuracy   <- mean(pred_class == test_data$High_Glucose)

cat(sprintf("Model accuracy: %.2f\n", accuracy))
Statistical Analysis Summary