Supplement 2. Python and WinBUGS model code for simulation study.

File List bugs_sim.py -- Python code to simulate data for input into WINBUGS. Program calls WINBUGS in a DOS command and sends summary output to a storage file. def_sim.py -- Python code (called by bugs_sim.py) specifying input parameters for simulations. oarunscript.txt -- WINBUGS script file for batch execution. oa_model.txt -- WINBUGS model file for batch execution. oa_data.txt -- WINBUGS data file (replaced at each simulation iteration). oa_inits.txt -- WINBUGS initial parameter value file (replaced at each simulation iteration). Description To execute program in WINDOWS operating systems, place all files in /Program Files/ WINBUGS14 and open the file "bugs_sim.py". This file references the "def_sim.py" and simulates data and initial values, replacing oa_data.txt and oa_intis.txt. The program then executes WINBUGS as a DOS command, with parameters contained in "oa_runscript.txt", using the simulated model specified in "oa_model.txt" and current data and initial values. Summary output is sent to two comma-delimited files: "sim_params.csv" and "sim_summary.csv", which keep track of parameter values and model estimates, respectively, for each simulation, and from which estimator performance (e.g., bias, MSE, and interval coverage) can be evaluated. <i>Note</i>: Execution of "bugs_sim.py" and "def_sim.py" requires installation of Python; we recommend ActiveState ActivePython 2.5 (http://www.activestate.com), and numpy numeric Python (http://www.scipy.org/Download). Installation of pymc (http://pymc.googlecode.com/files/pymc-2.0.win32-py2.5.exe) is required to access the likelihood objects in "bugs_sim.py"; alternatively users may code their own likelihood functions directly in Python.