R codes used for the paper "Conditional partial exchangeability: a probabilistic framework for multi-view clustering" by Franzolini, De Iorio, Eriksson.

Author contact: Beatrice Franzolini

This repository contains the R code used for simulations and data analysis in the manuscript and supplementary materials. The codes include the implementation of posterior MCMC for telescopic hierarchical Dirichlet processes (t-HDP) for different kernel choices and poly-tree structures.

Simulation_n1.R to run to get results of the simulation study for Scenario n.1 (results in Section 7.1 of the main paper and Section S5 of the Supplement) (The code can be sourced in full or executed line by line.)

Simulation_n2.R to run to get results of the simulation study for Scenario n.2 (results in Section 7.1 of the main paper and Section S5 of the Supplement) (The code can be sourced in full or executed line by line.)

Simulation_nA.R to run to get results of the simulation study for Scenario n.A (results in Section S5 of the Supplement) (The code can be sourced in full or executed line by line.)

Simulation_nB.R to run to get results of the simulation study for Scenario n.B (results in Section S5 of the Supplement) (The code can be sourced in full or executed line by line.)

Simulation_nC.R to run to get results of the simulation study for Scenario n.3 (The code can be sourced in full or executed line by line.)

Comparison_VariantwithDP.R to run to get the results corresponding to the DP-t-HDP model (and simulated data Scenario A. B. and 1) presented in Section S6 of the Supplement (The code can be sourced in full or executed line by line.)

Enriched_NNIG_uni.R to run to get the results corresponding to the enriched Dirichlet process with 10 layers normal kernel and Normal-Inverse-Gamma base measure (and simulated data Scenario n.1) (presented in Section 7.1 of paper and Section S5 and S6 of the Supplement) (The code can be sourced in full or executed line by line.)

main_metabolomics.R to run to get results on real data as presented in Section 7.2 of the main paper and Section S7 of the Supplement (real data and t-HDP model) WARNING: real data are not available, not all Figures are reproducible without real data!

HDP.R MCMC for a standard HDP (as in Teh et. al, 2006) Normal prior

telescopic_HDP_NN_uni.R MCMC for t-HDP Markovian Normal prior

telescopic_HDP_NN_uni_3L1P.R MCMC for t-HDP triangular Normal prior

telescopic_HDP_NN_uni_VariantwithDP MCMC for DP-t-HDP Markovian Normal prior

telescopic_HDP_NNIG_uni.R MCMC for t-HDP Markovian Normal-Inverse-Gamma prior

telescopic_HDP_NNIG_uni_VariantwithDP MCMC for DP-t-HDP Markovian Normal-Inverse-Gamma prior

telescopic_HDP_NNIW_multi.R MCMC for t-HDP Markovian Normal-Inverse-Wishart prior

telescopic_HDP_NNIW_multi_3L1P.R MCMC for t-HDP triangular Normal-Inverse-Wishart prior

telescopic_HDP_NNIX_multi.R MCMC for t-HDP Markovian Normal-Inverse-Chisquared prior

telescopic_HDP_NNIX_multi_3L1P.R MCMC for t-HDP triangular Normal-Inverse-Chisquared prior

Conditional_t-HDP.R code sourcing MCMC algorithms for the t-HDP models

functions_to_overwrite_to_extract_cluster_configurations_fromLSBP.R used in Simulation_n1.R to extract cluster configurations from LSBP-based models.

output directory containing csv files with MCMC chains for all models and data + a pdf with the codes used to map csv to each pair of data and model

Simulation_triangular.R to run to perform a simulation not used in the paper

toy_example_dimensionality.R to run to get Figure 1 with toy-example (Section 1 of the main paper) (The code can be sourced in full or executed line by line.)

For bug reporting purposes, e-mail Beatrice Franzolini (franzolini@pm.me).