This component contains the materials to conduct our sensitivity analysis.
We conducted this analysis using materials provided by the Center for High Throughput Computing (http://chtc.cs.wisc.edu/), which allows a user to send jobs from a single "submit node" out to thousands of computers simultaneously.
In this project, the following materials are required to send out these jobs:
1. **An instructions file** (a file with a .sub extension)
2. **job_info.txt**, which contains a job and seed number. These are read by the .sub file and sent to each of the submit nodes
3. **robust_conditions.csv**, a file with the same job numbers as in job_info.txt and the actual details of the specifications tested with that particular job number
4. **R.tar.gz**, a compressed version of R, which is sent to each of the submit nodes. This file is uncompressed on the submit node, and the installation of R is used to run ...
5. **sim_robust_v3.R**, which is contains the actual code for the sensitivity analysis. This file uses robust_conditions.csv to find the information on the models that need to be fit for the particular job number assigned to a particular computer to run
6. **A file with the .sh**, a basic shell script designed to run on a Linux machine that uncompresses R.tar.gz and runs the .R file
This component contains two sets of jobs: One that fits all the models in the specification curve (using the files nih_robust.sub and runR.sh) and one that conducts the permutation test to determine how the specification curve should perform under the null hypothesis (using the files nih_robust_resample.sub and runR_resample.sh).
The raw results are stored in **Sensitivity results.zip** and **Sensitivity results resample.zip**. These can be analyzed in R using the **check_specs.R** file.