Time-series Imputation Algorithm

Published

November 5, 2021

Author(s)

David A. Howe, Chloe Champagne

Abstract

Statistical imputation is a field of study that attempts to fill missing data. It is commonly applied to population statistics whose data have no correlation with running time. For a time series, data is typically analyzed using the autocorrelation, the Fourier transform to estimate power spectral density (PSD), trends, the Allan deviation (ADEV), and basically any analysis that depends on uniform time indexes. We describe an imputation algorithm that fills multiple gaps in a time series. We create an example in which four massive gaps exceed 100% of the original time series. We show that the PSD with imputation applied to the gaps is essentially the same as the original. Also, the confidence of ADEV with imputation falls within 90% of the original ADEV. The algorithm in Python is included for those wishing to try it.

Citation

IEEE Signal Processing Letters

Pub Type

Journals

Download Paper

https://doi.org/10.36227/techrxiv.16926694.v1

Local Download

Keywords

ADEV, convolution, dead-time, frequency, gaps, imputation, missing data, power-law noise models, Python, sparce, time, time-series.

Citation

Howe, D. and Champagne, C. (2021), Time-series Imputation Algorithm, IEEE Signal Processing Letters, [online], https://doi.org/10.36227/techrxiv.16926694.v1, https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=933257 (Accessed October 31, 2024)