-
Notifications
You must be signed in to change notification settings - Fork 47
CmdStan summary
Brian Lau edited this page May 16, 2017
·
3 revisions
Fits can be summarized using the print method of a StanFit object.
fit.print();
Inference for Stan model: eight_schools_model
4 chains: each with iter=(1000,1000,1000,1000); warmup=(0,0,0,0); thin=(1,1,1,1); 4000 iterations saved.
Warmup took (0.071, 0.075, 0.088, 0.079) seconds, 0.31 seconds total
Sampling took (0.12, 0.12, 0.11, 0.11) seconds, 0.45 seconds total
Mean MCSE StdDev 5% 50% 95% N_Eff N_Eff/s R_hat
lp__ -5.0e+00 7.7e-02 2.7 -9.6 -4.8e+00 -1.1 1188 2625 1.0e+00
accept_stat__ 8.6e-01 4.7e-03 0.21 0.33 9.5e-01 1.0 2038 4505 1.0e+00
stepsize__ 3.5e-01 1.2e-02 0.017 0.33 3.6e-01 0.37 2.0 4.4 3.1e+13
treedepth__ 3.4e+00 1.4e-02 0.53 3.0 3.0e+00 4.0 1384 3059 1.0e+00
n_leapfrog__ 1.2e+01 8.9e-02 5.2 7.0 1.5e+01 15 3481 7694 1.0e+00
divergent__ 0.0e+00 0.0e+00 0.00 0.00 0.0e+00 0.00 4000 8841 nan
energy__ 1.0e+01 1.0e-01 3.5 4.8 9.8e+00 16 1175 2597 1.0e+00
mu 8.0e+00 1.1e-01 5.1 -0.17 7.9e+00 16 2015 4455 1.0e+00
tau 6.4e+00 1.4e-01 5.4 0.44 5.2e+00 17 1522 3363 1.0e+00
eta[1] 3.7e-01 1.6e-02 0.94 -1.2 3.8e-01 1.9 3439 7600 1.0e+00
eta[2] -1.6e-02 1.4e-02 0.86 -1.5 2.4e-03 1.4 4000 8841 1.0e+00
eta[3] -1.9e-01 1.6e-02 0.94 -1.7 -1.9e-01 1.3 3599 7954 1.0e+00
eta[4] -5.4e-02 1.4e-02 0.91 -1.6 -6.0e-02 1.4 4000 8841 1.0e+00
eta[5] -3.8e-01 1.6e-02 0.88 -1.8 -4.0e-01 1.1 3109 6872 1.0e+00
eta[6] -2.0e-01 1.5e-02 0.90 -1.7 -1.9e-01 1.3 3546 7837 1.0e+00
eta[7] 3.4e-01 1.7e-02 0.92 -1.2 3.5e-01 1.8 3008 6649 1.0e+00
eta[8] 4.3e-02 1.6e-02 0.95 -1.5 4.8e-02 1.6 3605 7967 1.0e+00
theta[1] 1.1e+01 1.6e-01 8.2 -0.34 1.0e+01 27 2690 5946 1.0e+00
theta[2] 7.8e+00 9.6e-02 6.1 -2.0 7.8e+00 18 4000 8841 1.0e+00
theta[3] 6.3e+00 1.3e-01 7.6 -7.0 6.8e+00 18 3172 7012 1.0e+00
theta[4] 7.7e+00 1.1e-01 6.8 -3.6 7.7e+00 18 4000 8841 1.0e+00
theta[5] 5.1e+00 1.0e-01 6.4 -6.2 5.6e+00 15 4000 8841 1.0e+00
theta[6] 6.2e+00 1.0e-01 6.6 -5.2 6.7e+00 16 4000 8841 1.0e+00
theta[7] 1.1e+01 1.2e-01 6.9 0.59 9.9e+00 23 3085 6818 1.0e+00
theta[8] 8.4e+00 1.4e-01 7.8 -3.7 8.3e+00 21 2978 6582 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).
This summary is simply a capture of CmdStan's summary, but can be returned as a string array or a table:
[str,tab] = fit.print();
The string array mirrors the printed summary exactly, and the table tweaks the column names to conform to Matlab's naming conventions:
tab =
Mean MCSE StdDev p5_ p50_ p95_ N_Eff N_Eff_s R_hat
______ ______ ______ _____ ______ ____ _____ _______ _______
lp__ -5 0.077 2.7 -9.6 -4.8 -1.1 1188 2625 1
accept_stat__ 0.86 0.0047 0.21 0.33 0.95 1 2038 4505 1
stepsize__ 0.35 0.012 0.017 0.33 0.36 0.37 2 4.4 3.1e+13
treedepth__ 3.4 0.014 0.53 3 3 4 1384 3059 1
n_leapfrog__ 12 0.089 5.2 7 15 15 3481 7694 1
divergent__ 0 0 0 0 0 0 4000 8841 NaN
energy__ 10 0.1 3.5 4.8 9.8 16 1175 2597 1
mu 8 0.11 5.1 -0.17 7.9 16 2015 4455 1
tau 6.4 0.14 5.4 0.44 5.2 17 1522 3363 1
eta[1] 0.37 0.016 0.94 -1.2 0.38 1.9 3439 7600 1
eta[2] -0.016 0.014 0.86 -1.5 0.0024 1.4 4000 8841 1
eta[3] -0.19 0.016 0.94 -1.7 -0.19 1.3 3599 7954 1
eta[4] -0.054 0.014 0.91 -1.6 -0.06 1.4 4000 8841 1
eta[5] -0.38 0.016 0.88 -1.8 -0.4 1.1 3109 6872 1
eta[6] -0.2 0.015 0.9 -1.7 -0.19 1.3 3546 7837 1
eta[7] 0.34 0.017 0.92 -1.2 0.35 1.8 3008 6649 1
eta[8] 0.043 0.016 0.95 -1.5 0.048 1.6 3605 7967 1
theta[1] 11 0.16 8.2 -0.34 10 27 2690 5946 1
theta[2] 7.8 0.096 6.1 -2 7.8 18 4000 8841 1
theta[3] 6.3 0.13 7.6 -7 6.8 18 3172 7012 1
theta[4] 7.7 0.11 6.8 -3.6 7.7 18 4000 8841 1
theta[5] 5.1 0.1 6.4 -6.2 5.6 15 4000 8841 1
theta[6] 6.2 0.1 6.6 -5.2 6.7 16 4000 8841 1
theta[7] 11 0.12 6.9 0.59 9.9 23 3085 6818 1
theta[8] 8.4 0.14 7.8 -3.7 8.3 21 2978 6582 1