G10 FX Analysis

A statistical study of G10 foreign exchange return dynamics written in R.

Two years of daily spot rates are pulled from the Frankfurter API. The notebook then runs a set of classic quantitative checks: correlation structure, return distributions, rolling Sharpe ratios, and volatility clustering.

Key Questions Explored

• Which G10 currencies move together? How strong are the cross-currency correlations?
• How fat are the tails compared with a Normal distribution?
• How stable is the risk-adjusted return over rolling 30-day windows?
• Is there evidence of volatility clustering (a precondition for GARCH models)?

What the Notebook Produces

Section	Output
Summary statistics	Annualised return, vol, Sharpe, skewness, excess kurtosis per currency
Correlation heatmap	Full-period pairwise Pearson correlations, hierarchically ordered
Return distributions	Per-currency histograms with Normal fit overlay
Rolling Sharpe	30-day annualised Sharpe vs USD for each currency
Volatility clustering	Absolute returns over time with LOESS trend

Running It

Install dependencies once:

install.packages(c(
  "jsonlite", "dplyr", "tidyr", "ggplot2",
  "purrr", "lubridate", "patchwork", "scales", "zoo"
))

Then knit to GitHub-flavoured Markdown (creates analysis.md with embedded charts):

rmarkdown::render("analysis.Rmd")

Or open analysis.Rmd in RStudio and press Knit.

Requires R 4.1+. (everything is fetched live)

Tech Stack

• R 4.1+
• tidyverse (dplyr, ggplot2, tidyr, purrr)
• jsonlite for API calls
• zoo for rolling-window calculations
• patchwork and scales for plot composition
• lubridate for date arithmetic

Project Structure

g10-fx-analysis/
├── analysis.Rmd          # Main notebook (source)
├── analysis.md           # Rendered output with embedded charts (auto-generated)
└── .github/
    └── workflows/
        └── render.yml    # GitHub Action: re-renders on every push to main

Disclaimer

For educational purposes only. Nothing here constitutes financial advice.

---

© 2026 · Made by NavyBlueCheese