independent-clusterings-code

Code to reproduce simulations and data application from Gao, L.L., Bien, J. and Witten, W. (2018+) Are Clusterings of Multiple Data Views Independent?

Required R libraries:

• mclust
• matrixStats
• multiviewtest
• ggplot2
• grid
• gridExtra
• scales
• impute

Code Dependencies:

Code for some figures (e.g. Figures 2 and 3 in the main text) depend on the results of auxiliary files. Before running code for the figures, first:

• Run sim-study-sec4-sec5.R with p = 10 and all combinations of (delta, n, seed, K, sig) in {0, 0.125, 0.250, 0.375, 0.500, 0.625, 0.750, 0.875, 1} x {25, 50, 100, 250, 500} x {1, 2, ..., 12} x {3, 6} x {2.4, 4.8, 9.6}
• Run sim-study-sec4-sec5.R with p = 100 and all combinations of (delta, n, seed, K, sig) in {0, 0.125, 0.250, 0.375, 0.500, 0.625, 0.750, 0.875, 1} x {25, 50, 100, 250, 500} x {1, 2, ..., 12} x {3, 6} x {4.8, 9.6, 19.2}
• Run sim-study-sec4-sec5-collate-results.R with all combinations of (p, K) in {10, 100} x {3, 6}
• Run chooseK-setting1-sim-code.R, chooseK-setting2-sim-code.R, dense-diagonal-gaussian-sim-code.R, dense-gaussian-sim-code.R, and dense-t-sim-code.R with all combinations of (delta, n, seed) in {0, 0.125, 0.250, 0.375, 0.500, 0.625, 0.750, 0.875, 1} x {25, 50, 100, 250, 500} x {1, 2, ..., 12}

Data Dependencies:

Figure 4 and Table 1 in the main text use data from the Pioneer 100 Wellness Project (Price et. al., 2017), which is publicly available on dbGAP. After applying for access to the dbGAP data set and subsequent download of the data, Figure4+Table1.R can be directly applied to participant_data.csv in the main folder, and proteomics_DS.txt, clinical_labs_DS.txt, and metabolites_DS.txt in the dbgap_data subfolder.