pika - Evaluate Correlation Between Time Series • pika

Overview

pika is a lightweight R package for evaluating correlation between time series. It was developed at the MRC Centre for Global Infectious Disease Analysis for epidemiological applications, particularly for studying relationships between reproduction numbers (Rt) and population mobility during infectious disease outbreaks.

This package was developed as part of the research described in:

Ainslie KEC, Walters CE, Fu H, et al. Evidence of initial success for China exiting COVID-19 social distancing policy after achieving containment. Wellcome Open Research. 2020;5:81. doi: 10.12688/wellcomeopenres.15843.2

Key Features

Cross-correlation analysis: Determine the optimal lag between two time series where correlation is highest
Rolling correlation: Calculate time-varying correlation between paired time series
Reproduction number estimation: Wrapper around EpiEstim for group-wise Rt estimation
Percent change calculation: Convert count variables to percent change relative to a baseline period
Visualization tools: Plot time series with correlation overlays and lag distributions

Installation

You can install the development version of pika from GitHub:

# install.packages("remotes")
remotes::install_github("mrc-ide/pika")

Integration with orderly

To use pika within an orderly task, install from the drat repository:

drat:::add("ncov-ic")
install.packages("pika")

Quick Start

library(pika)

# Load example data
data(china_case_data)
data(exante_movement_data)

# Estimate reproduction number by province
rt_estimates <- estimate_rt(
  dat = china_case_data,
  grp_var = "province",
  date_var = "date",
  incidence_var = "cases"
)

# Join Rt estimates with mobility data
data_joined <- dplyr::left_join(
  rt_estimates,
  exante_movement_data,
  by = c("date_end" = "date", "province")
)

# Determine optimal lag between Rt and movement
lags <- cross_corr(
  dat = data_joined,
  date_var = "date_end",
  grp_var = "province",
  x_var = "r_mean",
  y_var = "movement",
  max_lag = 10
)

# Calculate rolling correlation
data_corr <- rolling_corr(
  dat = data_joined,
  date_var = "date_end",
  grp_var = "province",
  x_var = "r_mean",
  y_var = "movement",
  n = 14
)

# Visualize results
plot_corr(
  dat = data_corr,
  date_var = "date_end",
  grp_var = "province",
  x_var = "r_mean",
  y_var = "movement"
)

Functions

Function	Description
`cross_corr()`	Determine the lag at which cross-correlation is highest between two time series
`rolling_corr()`	Calculate rolling (moving) correlation between two time series
`estimate_rt()`	Estimate effective reproduction number (Rt) by group using EpiEstim
`calc_percent_change()`	Convert counts to percent change relative to a baseline period
`plot_corr()`	Visualize time series and rolling correlation
`plot_lag()`	Plot histogram of lags by group

Included Data

pika includes example datasets from the COVID-19 pandemic in China:

china_case_data: Daily confirmed COVID-19 cases by province (Jan 16 - Mar 24, 2020)
exante_movement_data: Daily within-city movement index by province (Jan 1 - Mar 24, 2020)

Documentation

For detailed usage examples and methodology, see the package vignette:

vignette("pika_vignette", package = "pika")

Citation

If you use pika in your research, please cite:

Ainslie KEC, Walters CE, Fu H, et al. Evidence of initial success for China exiting COVID-19 social distancing policy after achieving containment. Wellcome Open Research. 2020;5:81. doi: 10.12688/wellcomeopenres.15843.2

License

MIT License. See LICENSE for details.

Contributing

Contributions are welcome! Please feel free to submit issues or pull requests on GitHub.