In February 2021, the ANU Online Summer School in Political Analysis was launched by the Australian National University School of Politics & International Relations. Nine Waseda students from the Graduate School of Economics and Graduate School of Political Science (including one undergraduate student who is scheduled to go on to the GSPS this year) participated in the program and took courses such as text analysis taught by an instructor involved in the development of statistical software R, Bayesian statistics and political survey design.
Course attendance from Waseda:
Foundations of Statistical Analysis in Political Science（Feb. 1-8）: 1
Bayesian Statistics for Politics（Feb. 15-18）: 4
Political Survey Design and Analysis（Feb. 15-18）: 2
Introduction to Political Text Analysis（Feb. 15-18）: 2
Political Survey Design and Analysis/ Dr. Jill Sheppard
CHAI, Siyuan (Peter Chai)
School of Political Science and Economics, 4th Grade
Research Interests: Comparative Politics; Asian Political Economy; Public Opinion
This advanced course in political science teaches about qualitative survey design approaches, survey implementation, and quantitative survey response analysis using R programming. Through participating in this summer school, I was able to obtain updated insights into the hands-on skills and considerations related to sampling and response rate maximization strategies as well as data collection and cleaning procedures. This fruitful and invaluable experience of learning from the skillful course instructor, Dr. Jill Sheppard, who is an investigator of the Australian Election Study (AES), could enable me to utilize and apply the comprehensive knowledge I acquired from the course material and contents and the in-class exercises to my survey research planning and administration as a graduate student in the near future.
This advanced course is composed of four real-time lectures and four RStudio lab workshops and is an appropriate preparatory training for those who would like to undertake studies surrounding political science topics such as public opinion, voting behavior, political participation, and political culture. The course instructor elaborated on quintessential concepts and methodologies regarding survey studies in detail spanning from population coverage and sampling frame in the beginning, to collection mode and respondent recruitment for implementation, and to descriptive statistical outcomes and visualizations using “tidyverse,” “ggplot,” and “survey” packages at the end. Furthermore, she emphasized on external validity and internal reliability as well as a range of potential and common errors and biases that give rise to dilemmas and tradeoffs in survey design that need to be addressed by post-stratification weighting.
Applying “survey” Package in Rstudio
The participants of this course could easily refer to and employ the useful “Total Survey Error” framework taught in this summer school for their future research themes and target populations of focus. Therefore, based on my own experience, I strongly recommend this advanced course to fourth-year undergraduate students as well as graduate students who are endeavoring to become an all-round survey researcher to administer primary survey projects or to inspect available secondary survey datasets. This course would surely fill some knowledge gaps in the future participants’ research techniques!
Lastly, I would like to express my sincere gratitude to the Center for Positive/Empirical Analysis of Political Economy and the Top Global University project for subsidizing and supporting my eye-opening journey during this ANU SSPA which has brought me a new step further toward my graduate school-level research.
Introduction to Political Text Analysis/ Prof. Kenneth Benoit
LEUNG, Hoi Ki
Graduate School of Political Science, M1
Research Interests: Comparative politics
This is an introductory class on political text analysis. Prof. Kenneth Benoit is a frontier scholar in political text analysis, and he co-developed one of the most famous R package on text analysis “Quanteda”. He taught us fundamental ideas and practical practices on the topic and the application of “Quanteda”. For example, the R workflow of dealing with an original text: import text, transforming it into text units, statistical analyses, and visualization using word cloud, keyness graph.
Prof. Benoit is both very friendly and professional. He welcomes problems from every student and demonstrate ways of solving the problems instantaneously in class. For instance, when a student sent a pdf to him and asked for ways to extract text in a certain format, he solved it in class and demonstrated to other students right away, following up by sending the code to us. I also asked a few questions of my personal research project after the course finished and received generous advice from him.
Although interaction with other students was limited due to the online class nature, I would definitely and highly recommend this course for students who want an introduction to text analysis and “Quanteda”. To future students: Don’t be shy to ask problems in class! It is also a great learning opportunity to bring along practical questions you have and ask for Prof. Benoit’s opinion. As this is an introductory class, he has an awareness to make the course language simple and understandable by a complete novice in the field. Of course, it would help if you know some R in advance, but no need to worry if you know nothing of the R language or text analysis beforehand.
Bayesian Statistics for Politics/ Dr. Shawn Treier
Graduate School of Political Science, D3
Research Interests: Comparative politics
In this course, I learned the basics of Bayesian analysis. The course started with the basics of probability theory (e.g., conditional probability and Bayes rule), then moving on to theoretical issues of single/multiple parameter models. It also covered post-analysis issues such as regression diagnostics and prediction. The latter half of this course focused on more complicated topics such as factor analysis, hierarchical models, and logit/probit. After the course, participants must finish four assignments, which are mainly computational, for a certificate.
Although introductory, this course implicitly requires extensive knowledge of conventional frequentist statistics: probability distributions, hypothesis tests, matrix algebra, linear/non-linear regressions, maximum likelihood estimation, factor analysis, and so on. I managed to follow through the entire course, but I recommend that prospective participants have considerable knowledge of statistics before taking this course. Also, taking a look at some textbooks on Bayesian statistics (e.g., Greenberg, 2008) and exposure to the R language (not Stata) will definitely be useful.
By and large, this course was satisfactory. I was able to not only obtain a basic knowledge of the Bayesian framework but also revise my perspectives on statistics as a whole. Because I have stuck to the traditional approach (frequentist statistics), it was meaningful to learn the new approach, Bayesian statistics. However, there are some problems. First, I felt this 4-day course too short to understand Bayesian statistics fully. The explanation on theoretical aspects proceeded at lightning speed, and the instructor did not spare time to focus on details. Second, because it was held online, this course did not facilitate participants with each other. If it had been held on-site, I could have become familiar with other participants.