Western University Political ScienceFaculty of Social Science

Dave Armstrong

Associate Professor


PhD, University of Maryland
Telephone: 519.661.2111 ext. 85160
E-mail: dave.armstrong@uwo.ca
Office: Social Sciences Centre 4142

Research Interests

Professor Armstrong specializes in statistics and data analysis.  His research spans topics from measurement and latent trait estimation to the role of non-linearity and data mining techniques in statistical models.  

Current Research Projects

1. Non-linearity in Statistical Models

Observational studies use control variables to eliminate alternative explanations of the phenomenon under consideration. Often times, we treat control variables with less care and rigor in our model specifications than we do our own variables of interest.  This project considers how much attention we should pay to non-linearity in this context and what difference it might make for the effect sizes on variables of theoretical importance.

2. Costs of Contention

The Costs of Contention project seeks to understand how diverse forms of political conflict and violence (e.g. genocide, civil war, human rights violations) influence diverse political and economic outcomes (e.g., the type of political system, mass participation, economic development, happiness and foreign direct investment).

​Previous research relevant to this topic has been limited to studying only specific forms of conflict and violence as well as specific outcomes. The current project therefore opens up these categories to achieve a comprehensive analysis of the real costs of contention. Additionally, the project seeks to explore not only global patterns but also sub-national and individual level patterns. The research effort is complex in that it involves using pre-existing data in new ways, as well as collecting and analyzing new data across time and at multiple levels of analysis (i.e., the globe, specific country cases, and individual level data from specifically-targeted matched locations within the cases). The project will be attentive to numerous potential biases: e.g., gender differences in costs of contention, and the existence of missing data. The information emerging from this research will have significant potential use. Most importantly, project outputs will provide evidence-based early warning of likely challenges for recovery and development efforts in the post-conflict period, for policy-makers, practitioners, and other stakeholders engaged in recovery efforts. This can significantly improve the lives of those who would otherwise suffer without informed policy options.

3. Visualization of Pairwise Comparisons

There are many situations where we want to compare many pairs of statistical estimates.  For example, when a categorical independent variable is used in a regression model or when the outcome is a categorical variable with more than two values.  In these situations, there are many alternatives for visualizing these comparisons.  Each method seems to arise in virtual isolation (not acknowledging that others exist) which has led to a situation where no set of best practices exist.  This project looks to figure out what best practices ought to be and extend some of these ideas to dynamic settings through Shiny and D3.js applications. 

Selected Publications


Refereed Journal Articles

Recent Conference Presentations

Awards and Distinctions