Martin-Quinn Scores : Replication

[Updated July 7, 2009] It is now possible to fit the dynamic item response theory model using the Markov chain Monte Carlo algorithm discussed in the Political Analysis article using MCMCpack. The function MCMCdynamicIRT1d() fits the model; please consult the MCMCpack documentation for specifics. We are using this code to produce the scores released annually on the website. All data come for the Supreme Court Datbase. Because estimation takes place using simulation, it is likely you will get slightly different results between runs and across different platforms. Also, make sure to assess convergence before using the posterior density samples for inference.

To facilitate replication of the results published in Political Analysis, we provide these two tarballs:

Constant Ideal Point Model (constant.tar.gz)
Dynamic Ideal Point Model (dynamic.tar.gz)

The code uses the out-dated Scythe Statistical Library, Version 0.2, which is included in the tar-balls. We have compiled and run this code on Linux and MacOS X workstations using gcc 2.95. To compile and run this code on your system, it is necessary to change the paths in the code as well as in the Makefile. The data in the archives comes from the Supreme Court Data Base. All recodes from the raw data are documented in the article. Because estimation takes place using simulation, it is likely you will get slightly different results between runs and across different platforms. Also, make sure to assess convergence before using the posterior density samples for inference.