Data and Code for "The paradox of competition: How funding models could undermine the uptake of data sharing practices"

This repository contains simulation code, simulation output and analysis code and output for the above preprint.

The model was coded in NetLogo, and is available from the file data_sharing_policies.nlogo.

Code for the generation of the networks is available in the self-contained notebook network_generation/00-network-generation.qmd.

The analysis pipeline consisted of multiple steps:

1. Run the model using the scripts in batch_commands.
2. Pre-process the output files from the simulation to prepare them for analysis in Spark (files 01-move-to-parquet.R and 02-pivot-columns.R in pre-process).
3. Analyse various parts of the model with the analysis notebooks in analysis.

Due to size constraints we share the outputs from steps (1) and (3). The intermediate files from step (2) are larger than the files from step (1), and contain the same content.

A short note on naming conventions. In the paper, we speak of four types of networks, but the names are slightly different than in the code. This is just because the naming became more precise over the course of the analysis. The mapping between the output files and the reported networks is as follows:

1. "vary_incentives.csv.bz2" = No network.
2. "vary_incentives_individuals_clustered.csv.bz2" = high clustering.
3. "vary_incentives_individuals_fragmented.csv.bz2" = low clustering.
4. "vary_incentives_individuals_random network.csv.bz2" = random network.

Sensitivity analysis

The repo also contains outputs and analysis notebooks for the sensitivity analysis. The analysis was done in Spark due to the large file sizes. We share three "packages" of data that we used for the sensitivity analysis:

• gain-sensitivity-data.tar.bz2
• sigma-sensitivity-data.tar.bz2
• costs-sensitivity-data.tar.bz2

These archives contain two files each: a general file for the sensitivity analysis, and one with individual-level data. Both are stored as .parquet files.

They are already processed (similar to step 2 above), so can be readily analysed using the files 10-Figure-1-sensitivity.qmd, 10-Figure-2-sensitivity.qmd, and 10-Figure-3-sensitivity.qmd which are available under analysis.