The prompts used as examples in the slides from the lab on "AI for Coding".

# Example prompts and answers

The prompts used as examples in the slides from the lab on "AI for Coding".

Results were created by Claude Sonnet 3.7 in May 2025.

code_documentation.R

# Results were created by Claude Sonnet 3.7 in May 2025

# Prompt:---------------

# I have written the following function that computes ROI in R:
#   
#   calculate_roi <- function(revenue, cost) { 
#     roi <- (revenue - cost) / cost * 100 return(roi) 
#   }
#   
#   Please provide a full roxygen2 documentation of the function, including an example.

# Result:---------------

#' Calculate Return on Investment (ROI)
#'
#' @description
#' Calculates the return on investment as a percentage based on revenue and cost inputs.
#' ROI measures the profitability of an investment relative to its cost.
#'
#' @param revenue Numeric. The total revenue or return generated from the investment.
#' @param cost Numeric. The initial cost or investment amount. Must be non-zero.
#'
#' @return Numeric. The ROI expressed as a percentage.
#'
#' @details
#' The function uses the standard ROI formula: ((revenue - cost) / cost) * 100
#' A positive result indicates profit, while a negative result indicates loss.
#'
#' @examples
#' # Basic example
#' calculate_roi(150000, 50000)  # Returns 200
#'
#' # Marketing campaign example
#' campaign_revenue <- 75000
#' campaign_cost <- 25000
#' campaign_roi <- calculate_roi(campaign_revenue, campaign_cost)  # Returns 200
#'
#' # Investment loss example
#' calculate_roi(8000, 10000)  # Returns -20
#'
#' @export
calculate_roi <- function(revenue, cost) {
  roi <- (revenue - cost) / cost * 100
  return(roi)
}

code_explanation.R

# Prompt:---------------

# I am operating in a management context. I use the sales data from above. 
# I want to predict future sales using Meta's prophet algorithm. I found the 
# following R code that does this. I do not understand this code. 
# Please explain it to me step by step.

library(prophet)
library(dplyr)

sales_forecast <- function(data) {
  # Rename columns to Prophet's required format
  df <- data %>%
    rename(ds = quarter_date, y = sales) %>%
    select(ds, y)
  
  # Create Prophet model with quarterly seasonality
  m <- prophet(df, 
               seasonality.mode = 'multiplicative',
               yearly.seasonality = TRUE,
               weekly.seasonality = FALSE,
               daily.seasonality = FALSE)
  
  # Add quarterly seasonality
  m <- add_seasonality(m, name = 'quarterly', period = 365.25/4, 
                       fourier.order = 5)
  
  # Create future dataframe for 8 quarters ahead
  future <- make_future_dataframe(m, periods = 8, freq = 'quarter')
  
  # Make forecast
  forecast <- predict(m, future)
  
  # Return results with components
  return(list(model = m, 
              forecast = forecast, 
              plot = plot(m, forecast),
              components = prophet_plot_components(m, forecast)))
}

sales_forecast(sales_clean)

# Example 6: debugging-----------

library(dplyr)

df <- data.frame(id = 1:3, group = factor(c(1, 2, 3)))

filtered_df <- df %>% 
  filter(group > 1)

print(filtered_df)

code_generation.R

# Results were created by Claude Sonnet 3.7 in May 2025

# Prompt:---------------
# Help me in writing R code. Stay within the tidyverse dialect and explain your code using inline comments.
# 
# I have a data set on wine:
#   alcohol quality residual sugar  kind
# 1     8.8       6           20.7 white
# 
# Write code that computes average alcohol for all kinds of wine.

# Result:---------------
library(dplyr)
wine_data <- DataScienceExercises::wine2dine  # <- manually added to replace Claude's sample data set
wine_averages <- wine_data %>%
  group_by(kind) %>%                    # Group the data by wine type
  summarize(
    avg_alcohol = mean(alcohol),        # Calculate mean alcohol content
    count = n()                         # Count number of wines in each group
  )

syntax_support.R

# Results were created by Claude Sonnet 3.7 in May 2025

# Prompt:---------------
# I need to analyze our company's quarterly sales data. I want to do this using R and the tidyverse dialect. But my background is in management studies, not programming. Please provide advice on how I can analyze the data to identify trends and seasonality.
# 
# I have the data in the form of a csv file that has the following structure:
# "quarter_date","quarter","sales","region","product_category"
# 2020-01-01,"Q1 2020",12659.86530177,"North","Furniture"
# 
# Write R core that:
# * Imports the csv file
# * Cleans the data
# * Create a figure using ggplot2 that highlights trends and seasonality

# Result:---------------
# Step 1: Load necessary packages
library(readr)       # For reading CSV files
library(dplyr)       # For data manipulation
library(tidyr)       # For data cleaning
library(ggplot2)     # For data visualization
library(lubridate)   # For handling dates
library(ggpubr)      # For arranging multiple plots

# Step 2: Import the CSV file
# Replace "quarterly_sales.csv" with your actual file path if different
sales_data <- read_csv("sales_data.csv")

# Step 3: Examine the data structure
head(sales_data)  # View first few rows
summary(sales_data)  # Get summary statistics

# Step 4: Clean the data
sales_clean <- sales_data %>%
  # Ensure quarter_date is properly formatted as date
  mutate(quarter_date = as.Date(quarter_date)) %>%
  # Handle missing values in sales column by replacing with mean
  mutate(sales = if_else(is.na(sales), 
                         mean(sales, na.rm = TRUE),
                         sales))

# Step 5: Create visualizations for trends and seasonality

# 5.1: Overall sales trend over time
trend_plot <- ggplot(sales_clean, aes(x = quarter_date, y = sales)) +
  geom_line(color = "blue", size = 1) +
  geom_point(color = "darkblue", size = 3) +
  geom_smooth(method = "loess", se = FALSE, color = "red", linetype = "dashed") +
  labs(title = "Quarterly Sales Trend (2020-2024)",
       x = "Quarter",
       y = "Sales ($)") +
  theme_minimal() +
  scale_y_continuous(labels = scales::dollar_format()) +
  scale_x_date(date_breaks = "6 months", date_labels = "%b %Y")

# 5.2: Seasonal patterns by extracting quarter
seasonal_plot <- sales_clean %>%
  # Extract quarter number (1-4) from date
  mutate(quarter_num = lubridate::quarter(quarter_date)) %>%
  # Group by quarter to see seasonal patterns
  group_by(quarter_num) %>%
  summarize(avg_sales = mean(sales, na.rm = TRUE)) %>%
  ggplot(aes(x = factor(quarter_num), y = avg_sales)) +
  geom_col(fill = "skyblue", width = 0.6) +
  geom_text(aes(label = scales::dollar(avg_sales, accuracy = 1)), 
            vjust = -0.5, size = 3.5) +
  labs(title = "Average Sales by Quarter",
       x = "Quarter",
       y = "Average Sales ($)") +
  theme_minimal() +
  scale_y_continuous(labels = scales::dollar_format())

# 5.3: Regional comparison
region_plot <- sales_clean %>%
  group_by(region) %>%
  summarize(total_sales = sum(sales, na.rm = TRUE)) %>%
  ggplot(aes(x = reorder(region, -total_sales), y = total_sales)) +
  geom_col(fill = "lightgreen") +
  labs(title = "Total Sales by Region",
       x = "Region",
       y = "Total Sales ($)") +
  theme_minimal() +
  scale_y_continuous(labels = scales::dollar_format())

# Step 6: Arrange all plots in one figure with ggpubr::ggarrange()
combined_plot <- ggpubr::ggarrange(
  trend_plot, seasonal_plot, region_plot,
  ncol = 3,             # Stack horizontally
  nrow = 1,             # Three rows
  heights = c(1.2, 1, 1),  # Slightly larger height for trend plot
  labels = c("A", "B", "C"),  # Add labels to each subplot
  align = "v"           # Align plots vertically
)

# Add a common title to the combined figure
final_plot <- ggpubr::annotate_figure(combined_plot,
                                      top = ggpubr::text_grob("Quarterly Sales Analysis (2020-2024)", 
                                                              face = "bold", size = 16),
                                      bottom = ggpubr::text_grob("Source: Company Sales Data", 
                                                                 hjust = 1, x = 1, face = "italic", size = 10)
)

# Step 7: Save the combined figure as PDF
ggpubr::ggexport(
  final_plot,
  filename = "sales_analysis.pdf",
  width = 16,
  height = 5,
  dpi = 300
)

# Alternatively, you can use ggsave, but ggexport is preferred for ggpubr objects
# ggsave("sales_analysis.pdf", final_plot, width = 10, height = 14)

# Print the combined figure to screen
print(final_plot)

syntax_support_create_sales_data.R

library(tibble) # For tibble functionality
library(dplyr) # For data manipulation
library(lubridate) # For date handling
library(ggplot2) # For visualization

# Set seed for reproducibility
set.seed(123)

# Create base sales values first
quarter_dates <- seq(as.Date("2020-01-01"), as.Date("2024-10-01"), by = "quarter")
n_quarters <- length(quarter_dates)

# Calculate the sales values with trend, seasonality and noise
base_sales <- 100000 + 
  (1:n_quarters * 5000) +               # Upward trend
  (sin(1:n_quarters * pi/2) * 30000) +  # Seasonal pattern
  rnorm(n_quarters, 0, 15000)           # Random noise

# Randomly select some positions to replace with NA
missing_positions <- runif(n_quarters) > 0.9
base_sales[missing_positions] <- NA_real_ # missing double vector

# Now create the tibble with the pre-calculated sales values
sales_data <- tibble(
  # Create quarterly dates
  quarter_date = quarter_dates,
  
  # Create quarter labels (Q1 2020, Q2 2020, etc.)
  # Using lubridate explicitly with the quarter function
  quarter = paste0("Q", lubridate::quarter(quarter_date), " ", lubridate::year(quarter_date)),
  
  # Use the pre-calculated sales values
  sales = base_sales,
  
  # Add a region column
  region = sample(c("North", "South", "East", "West"), n_quarters, replace = TRUE),
  
  # Add a product category
  product_category = sample(c("Electronics", "Furniture", "Clothing", "Office Supplies"),
                            n_quarters, replace = TRUE)
)

# Preview the data
head(sales_data)

write.csv(sales_data, "sales_data.csv", row.names = FALSE)