This repo includes code to accompany our manuscript, “Efficient and Scalable Bipartite Matching through Fast Beta Linkage (fabl)”. The required functions are stored in the R folder, and they can be loaded by selecting Build -> Load All from the top menu of RStudio.

In particular;

• gibbs_base is used to run the gibbs sampler as described in Section 3
• hash is used to conduct the hashing and computation of $\tilde{\Gamma}$ as described in Section 4.1.
• gibbs_efficient is used to run the gibbs sampler as described in Section 4.2
• combine_hash is used to synthesize summary statistics from multiple batches of comparison vectors, as described in Section 4.3
• sei implements Storage Efficient Indexing, as described in Section 4.4.

The accuracy simulation from Section 5.2 can be found in code/sadinle_sim.R. The speed simulations from Section 5.1 cab be found in code/speed_big.R and code/speed_big2.R. After running this code, plots can be generated through files found in the local folder.

Data for the National Long Term Care Study (NLTCS) and El Salvador human rights data are restricted, and therefore not included here.