On September 26, I attended the Quantitative Methods & Reasoning Across Disciplines workshop hosted by the Quantitative Research Consulting Center and the Interactive Technology and Certificate Program as part of their Fall Skills Lab series. The two hour session covered topics in statistical analysis, data collection, and research design to empower an the group to incorporate quantitative approaches in the own projects.
The workshop served several purposes, the first being a refresher in fundamental and more intermediate topics in Statistics, while the rest of the time was spent with an eye toward applying those methods in research. In the first, and simplest portion on Statistics, we discussed measures of centrality in a dataset (e.g., Mean, Median, and Mode), as well as ranges within a dataset the as it relates to skew and volatility. I was familiar with these concepts already from undergraduate Statistics.
From there, we discussed measures of correlation and when it was appropriate to use each one. The Pearson Correlation Coefficient and Linear Regression both describe the relationships between two variables. I remembered both from Algebra and Statistics at one point or another, but was happy to think about them in the context of my interests within the Digital Humanities. The Pearson Correlation Coefficient is a value ranging from 0 to 1 or -1, indicating the strength of the relationship in either a positive (i.e., change in one variable leads to a change in the same direction of the other) or negative (i.e., change in one variable leads to a change in the opposite direction of the other) direction. Linear Regression also describes how one variable impacts another with the formula y = mx + b, but should be used when trying to predict another value based on one you already have. I was much less comfortable with the forms of Supervised and Unsupervised Classification we spoke about next.
Logistic regression is a supervised classification model used when your dataset contains two possible values (e.g., whether or not someone is employed job). The model is supervised because it is trained on a dataset. K means clustering is an unsupervised classification scheme that breaks data down into identifiable groups by partitioning the set down into clusters organized around the nearest mean. It is considered unsupervised because the algorithm learns patterns without being trained.
We concluded the workshop by discussing research design, touching on randomized experiments, natural experiments, and survey analysis. The last two, natural experiments, which are more so observational studies, and surveys seemed more apt for Digital Humanities subjects, it was helpful to understand how research is carried out to better understand academic works we read, as well.
I found the session to be a great resource and a really strong grounding in quantitative social science methods that I will keep in mined while pursuing my own projects and research within DH.
