Hi, I'm trying to figure out how to convert, in the basic_estimators bernoulli example, theta (and the gradient) in the bernoulli_samples.csv and bernoulli_diagnostics.csv files as produced by:
run(`./bernoulli sample random seed=1 data file=bernoulli.data.R output file=bernoulli_samples.csv diagnostic_file=bernoulli_diagnostics.csv`). Thanks a lot, Rob J. Goedman [email protected] Warmup took (0.011) seconds, 0.011 seconds total Sampling took (0.034) seconds, 0.034 seconds total Mean MCSE StdDev 5% 50% 95% N_Eff N_Eff/s R_hat lp__ -8.8 3.5e-02 7.4e-01 -10 -8.5 -8.3 441 13124 1.0e+00 accept_stat__ 0.90 4.7e-03 1.5e-01 0.58 0.96 1.0 1000 29734 1.0e+00 stepsize__ 1.1 6.3e-15 4.4e-15 1.1 1.1 1.1 0.50 15 1.0e+00 treedepth__ 1.7 1.6e-02 4.7e-01 1.0 2.0 2.0 920 27347 1.0e+00 n_leapfrog__ 2.3 3.1e-02 9.5e-01 1.0 3.0 3.0 920 27347 1.0e+00 n_divergent__ 0.00 0.0e+00 0.0e+00 0.00 0.00 0.00 1000 29734 nan theta 0.50 6.6e-03 1.4e-01 0.26 0.51 0.72 440 13068 1.0e+00 Samples were drawn using hmc with nuts. For each parameter, N_Eff is a crude measure of effective sample size, and R_hat is the potential scale reduction factor on split chains (at convergence, R_hat=1). ---------------------------------------------------------------------------------------- Inference for Stan model: bernoulli_model 1 chains: each with iter=(1000); warmup=(0); thin=(1); 1000 iterations saved. Warmup took (0.011) seconds, 0.011 seconds total Sampling took (0.034) seconds, 0.034 seconds total Mean MCSE StdDev 5% 50% 95% N_Eff N_Eff/s R_hat lp__ -8.8e+00 3.5e-02 7.4e-01 -10 -8.5e+00 -8.3 441 13124 1.0e+00 accept_stat__ 9.0e-01 4.7e-03 1.5e-01 0.58 9.6e-01 1.0 1000 29734 1.0e+00 stepsize__ 1.1e+00 6.3e-15 4.4e-15 1.1 1.1e+00 1.1 0.50 15 1.0e+00 treedepth__ 1.7e+00 1.6e-02 4.7e-01 1.0 2.0e+00 2.0 920 27347 1.0e+00 n_leapfrog__ 2.3e+00 3.1e-02 9.5e-01 1.0 3.0e+00 3.0 920 27347 1.0e+00 n_divergent__ 0.0e+00 0.0e+00 0.0e+00 0.00 0.0e+00 0.00 1000 29734 nan theta -4.0e-03 2.9e-02 6.0e-01 -1.1 2.3e-02 0.96 441 13127 1.0e+00 p_theta -1.7e-02 5.0e-02 1.6e+00 -2.6 -8.4e-02 2.6 1000 29734 1.0e+00 g_theta -4.5e-04 7.9e-02 1.7e+00 -2.9 6.9e-02 2.7 440 13068 1.0e+00 Samples were drawn using hmc with nuts. For each parameter, N_Eff is a crude measure of effective sample size, and R_hat is the potential scale reduction factor on split chains (at convergence, R_hat=1). method = optimize optimize ... initial log joint probability = -8.16567 Iter log prob ||dx|| ||grad|| alpha # evals Notes 5 -6.93147 2.54998e-05 2.50688e-12 1 8 Optimization terminated normally: Convergence detected: change in objective function was below tolerance elapsed time: 0.003743127 seconds (2416 bytes allocated) julia> log(0.5) -0.6931471805599453 julia> 10*log(0.5) -6.931471805599453
