I’m starting a replication workshop with my statistics students. During my lectures in basic statistics I found that students were most engaged in lab sessions when we worked with data sets. I want to build on that and use replication as a new teaching tool, and I will report about this on my blog. My first challenge: How to pick a paper that’s simple enough for statistics beginners?
Most students that signaled interest in replication have done Bivariate Statistics in R with me last term, and just finished my class Regression in R. Since I’m new to this, I will mainly follow the guidelines for replication as an educational tool by political scientist Gary King at Harvard. I have opened a Dataverse (which right now includes a dummy example) that ideally would look like King’s dataverse some day.
This would be the first replication workshop in Cambridge Social Sciences, as far as I know, therefore I’m eager to get feedback by scholars who have done this – especially with statistics beginners. Right now, I have planned for 8 two-hour-sessions.
What papers to pick?
King suggests, if one starts with articles with “less advanced methods (such as only linear regression)” one can then select a “more advanced method that makes sense to use” to be able to “extract more information” from the data. Statistics modules at the Social Sciences Research Methods Centre at Cambridge are usually 4 weeks long. This means that the ‘cohort’ of students that I’ve been teaching have done statistics at a minimum in 16 hours of teaching, plus 2 hours introduction in R, plus hopefully many hours practicing for our weekly quizzes and the final exams. For this replication workshop we will need only articles with “less advanced methods” I would think.
The first step, it seems, is to find articles that have used bivariate statistics and OLS regression. Bivariate correlation tables can be found practically everywhere because of multicollinearity issues. I’m thinking of letting students reproduce these tables in R as a start. For OLS regression I was less successful in finding suitable studies yet – most articles I know of deal with panel data (time series cross section). Here, I need help urgently on how to find studies that can work for stats beginners. Ideally these studies would be fairly recent, otherwise I’m not sure if students see the point in looking at ‘old’ work.
So my strategy is to get help finding such articles on Facebook, twitter, and shamelessly use this blog to get comments and feedback. I’m also planning to follow anyone who has done replication with their students before – so please get in touch!
These are my current guidelines on how to pick a paper (based on Gary King):