autocorrstats helps you test relationships in autocorrelated time series without relying on methods that assume independent samples. It follows Ebisuzaki (1997) to generate ensembles of synthetic time series with power spectra similar to those of the original data, allowing you to estimate the significance of statistical quantities.
In this repository, the word surrogates refers to those synthetic time series. They are not new observations; they are randomized series generated from the original data so that key properties such as the power spectrum are preserved while the timing information is scrambled.
The package currently supports significance testing for correlations and polynomial fits in time series that exhibit autocorrelation.
The simplest setup is the included Conda environment:
conda env create -f environment.yml
conda activate autocorrstatsThis installs the package in editable mode together with the development and notebook dependencies used in the examples.
If you only want to install the package itself:
pip install git+https://github.com/anthony-meza/autocorrstats.git@mainautocorrstats was created by Anthony Meza and is released under the MIT License.
Ebisuzaki, W. (1997). A method to estimate the statistical significance of a correlation when the data are serially correlated. Journal of Climate, 10(9), 2147-2153. https://doi.org/10.1175/1520-0442(1997)010%3C2147:AMTETS%3E2.0.CO;2