Bi-directional Interrupted Time series analysis
Keywords
Loading...
Authors
Issue Date
2024-04-01
Language
en
Document type
Journal Title
Journal ISSN
Volume Title
Publisher
Title
ISSN
Volume
Issue
Startpage
Endpage
DOI
Abstract
Discontinuity analyses are statistical methods that can be used to evaluate the
effects of interventions or treatments by exploiting naturally occurring discontinuities
in data. It leverages the idea that just above and below a certain threshold exposure
to some intervention may differ, allowing for causal inference under certain
assumptions. This approach is particularly useful in quasi-experimental settings
where randomization is not feasible or ethical. Discontinuity analysis has found
applications in various fields including economics, political science, education, and
public health. No methods exist that allow for the analysis of two simultaneous
interventions. We propose a new method called BITS that enables the separation
of simultaneously occurring interruption effects. We do so by establishing forward
and backward counterfactuals and interpolating between them. In doing so we can
control for the longer-lasting effect. We analyze the methodology mathematically,
perform simulations on toy data and finally apply the method to air pollution data
from the Ukrainian front. Our results show that BITS can effectively separate
interruption effects for simple functions. In complex functions more the method
relies heavily on correct assumptions to yield interpretable results.
Air pollution poses significant dangers to public health. War zones in particular
are hypothesised to exacerbate the harmful effects on human health and
the environment. Discontinuity analysis, offers a promising approach to studying
the impacts of conflict on air pollution. We propose leveraging known start and
end dates of conflict in a novel methodology to measure the impact of conflict
on exposure levels. Existing methods do not deal with scenarios involving the simultaneous
occurrence of multiple interventions. Like the decrease in anthropic
activity coupled with the direct impact from modern warfare on the frontlines of the
Russo-Ukrainian conflict. We propose a novel approach called BITS (bi-directional
time series), that is designed to separate concurrent interruption effects. The BITS
method establishes forward and backward counterfactuals and interpolates between
them, enabling the isolation of short-term interruption effects while controlling for
longer-lasting changes. We present a mathematical analysis of the methodology,
conduct simulations using toy data, and apply the method to air pollution data
from the Ukrainian front. Our findings demonstrate that BITS effectively separates
interruption effects for simple functions, although its performance in complex
scenarios relies heavily on the accuracy of underlying assumptions.
Description
Citation
Supervisor
Faculty
Faculteit der Sociale Wetenschappen