CSDS Study Group Presentations: 2022

Autumn Quarter 2022

Stuart Lynn, PhD
October 13, 2022

Over the past year, we have been working on a suite of open source tools that we call Matico. These tools, which can each be used individually, collectively form a new, forever open spatial data science platform which enables users to manage, explore, analyze and build applications centered around spatial data with little or no code. In this talk, we will explore the features of Matico by building out a state specific application for covid data. We will invite the audience to follow along in the creation of this application and afterwards we will discuss our roadmap for the next year of development of the platform and facilitate a conversation about potential use cases and features.

Renan Serenini
Co-presented with Frantisek Masek 
October 18, 2022

We estimate the causal effect of the Russian invasion of Ukrainian regions Donetsk and Luhansk in 2014 on household disposable income. We are particularly interested in estimating both direct and indirect treatment effects. Hence, besides estimating economic consequences on the two regions, we also investigate the spillover effects on neighboring regions. Based on possible violation of the SUTVA assumption when applying traditional quasi-experimental methods and including different Ukrainian regions amongst control units, we deploy tools that handle potential spillover effects of the treatment. Specifically, we use the spatial difference-in-differences from Delgado and Florax (2015), an extension of the well-known Diff-in-diff method that controls for possible indirect spatially affected units. On top of that, we also apply an extension of the synthetic control method (SCM) - the inclusive SCM - from Stefano and Mellace (2020), which can account for affected units and estimate the spillover effect. We find strong evidence for the direct treatment effect, yet we cannot document any presence of spillovers through the indirect treatment effects.

Franktisek Masek
Co-presented with Renan Serenini
October 25, 2022

We introduce a spatial version of the Synthetic Difference-in-Differences estimator of Arkhangelsky et al. (2021). The extension of the estimator builds on the approach of Delgado and Florax (2015) in their Spatial Difference-in-Differences method. Hence, we assume the treatment effect has spillover characteristics on some neighboring units. In other words, we allow treatment to have direct and indirect effects. We incorporate the framework into Arkhangelsky et al. (2021) showing that Delgado and Florax (2015) is a special case absenting the synthetic control weighting component of the units in the donor pool. We compare the Spatial Synthetic Difference-in-Differences with the estimator of Arkhangelsky et al. (2021) using an example of a violation of the SUTVA assumption leading to the treatment effect estimator being biased and inconsistent. All the features presented in Arkhangelsky et al. (2021) related to the comparison of the Synthetic Difference-in-Differences and conventional Difference-in-Differences carry forward in our case.

Sara Lafia, PhD
October 25, 2022

Archives that provide access to high quality research data offer an ideal setting to study the community impact of data sharing. In this talk, I will describe the data citation network at ICPSR – a large social science data archive at the University of Michigan – which reflects patterns of data reuse in academic literature. I have detected 41 research communities that use the same scientific datasets, which I characterize as disciplinary "subdivisions" or as integrative "crossroads" connecting research areas. I show how spatial perspectives on data use offer insights into scholarly communication processes, such as the organization of scientific communities.

Lilly Cheng Immergluck, M.S.,  M.D., FAAP
November 8, 2022

This spatial study group will discuss the spatial aspects of a research project under development:

Dr. Immergluck is a pediatric infectious disease specialist and population health service researcher, who has been working on spatial statistical models to address antibiotic resistant infections which are derived from the community (pre-CoVID pandemic). She has collaborated with Drs. Lance Waller, George Rust, and the late Robert Daum over the past 15 years.  She is currently developing a new research project to better understand the factors (across space and time) that affect particular strains of antibiotic resistant bacteria to cause infections in children living in the southeastern U.S. 

Gang Xu, PhD
November 15, 2022

GIScience has been widely used in understanding this pandemic. This talk includes two case studies about the spatial pattern of the COVID-19 epidemic in China. The first case study is the Geographically varying relationships between population flows from Wuhan and COVID-19 cases in Chinese cities. Previous studies demonstrated that population outflows from Wuhan determined COVID-19 cases in other Chinese cities but neglected the spatial heterogeneities of their relationships. We used the Geographically Weighted Regression (GWR) model to investigate the spatially varying influences of outflows from Wuhan. The second case study is about Spatial disparities of self-reported COVID-19 cases and influencing factors in Wuhan, China. We collected nearly one thousand self-reported cases from a social media platform during the early stage of the COVID-19 epidemic in Wuhan, China. We adopted the GWR model to quantify the influences of population dynamics, transportation, and social interactions on self-reported COVID-19 cases in Wuhan.

Friedrich Weise
November 15, 2022

Geospatial data is of high relevance in modeling of hydrogen ecosystems, for instance to evaluate available areas, transport distances or existing infrastructure. Spatial clustering can play a key role to group this data in spatial entities depending on geo-technical conditions rather than administrative boundaries. Therefore, I evaluated multiple clustering algorithms with various variables and parameters to find the best spatial resolution fitting to our techno economic modeling done at Fraunhofer ISE. During the talk I will give insights to our applications, the selected data and methodology, my results, and indicators I developed to compare the clustering algorithms beyond the classic sum of squares.