# test input file for graph_sampler
# dynamic directed acyclic graph (DBN) inference with 3 nodes,
# first 10 iterations,
# autocycles allowed, tempered MCMC sampling,
# basic edge prior and edge count prior, normal-gamma data model and
# missing data
# =========================================================================

n_runs = 10;
n_burn_in = 0;

random_generator = mersenne_twister;

random_seed = 49740;

# perk is the inverse temperature used in tempered MCMC
perk_scale = array {0.1, 0.5, 0.9, 1};

n_nodes = 3;

dynamic_bayesian_network = true;

autocycle = true;

initial_adjacency = matrix {empty};

hyper_pB = matrix {equanimous};

edge_count_prior = 1;

likelihood = normal_gamma;

n_data = 4;

data = matrix {NaN, 1.2, 1.4, 1.35,
               2.1, NaN, 2.5, 2.45,
               3.4, 3.6, NaN, 3.85};

save_chain = true;
n_saved_adjacency = 10;
save_best_graph = true;  
save_edge_probabilies = true;

# End.
