stanr

stanr is an R interface for compiling Stan programs and running Stan services directly from R. It uses an R6-based API similar to cmdstanr: stan_model() returns a StanModel object, and inference methods ($sample(), $optimize(), $variational(), etc.) return R6 fit objects with methods for extracting draws, summaries, and diagnostics. Results use posterior draw objects where possible.

Inference Methods

Method	Purpose	Return type
`$sample()`	MCMC sampling (HMC/NUTS, static HMC, WALNUTS, fixed param)	`StanMCMC`
`$optimize()`	Posterior mode or maximum likelihood estimate	`StanMLE`
`$laplace()`	Laplace approximation draws around a mode	`StanLaplace`
`$variational()`	ADVI approximate posterior draws	`StanVB`
`$pathfinder()`	Pathfinder approximate posterior draws	`StanPathfinder`
`$generate_quantities()`	Generated quantities from existing draws	`StanGQ`
`$diagnose()`	Gradient checking diagnostic	`StanDiagnose`

Fit Object Methods

All fit objects inherit from StanFit and share common methods:

Method	Description
`$draws()`	Extract draws as `posterior` objects
`$summary()`	Summarize draws via `posterior::summarise_draws()`
`$print()`	Print a summary table
`$return_codes()`	Stan return codes (0 = success)
`$metadata()`	Fit metadata (seed, data, arguments)
`$time()`	Timing information
`$log_prob()`	Evaluate log probability
`$constrain_variables()`	Constrain unconstrained parameters
`$unconstrain_variables()`	Unconstrain parameters
`$save_object()`	Save fit to file

Installation

Install pre-built binaries from R-Universe via:

install.packages('stanr', repos = c('https://andrjohns.r-universe.dev', 'https://cloud.r-project.org'))

You can also install the development version of stanr from GitHub:

# install.packages("pak")
options(pkg.git_submodules = TRUE)
pak::pak("git::https://github.com/andrjohns/stanr")

Compile a Model

Models are compiled from a string or from a .stan file. stan_model() returns a StanModel R6 object.

library(stanr)

bernoulli_model <- "
data {
  int<lower=0> N;
  array[N] int<lower=0, upper=1> y;
}
parameters {
  real<lower=0, upper=1> theta;
}
model {
  theta ~ beta(1, 1);
  y ~ bernoulli(theta);
}
generated quantities {
  array[N] real log_lik;
  for (n in 1:N) {
    log_lik[n] = bernoulli_lpmf(y[n] | theta);
  }
}
"

mod <- stan_model(code = bernoulli_model, model_name = "bernoulli")
data <- list(N = 10, y = c(1, 0, 1, 1, 0, 1, 0, 0, 1, 0))

Sampling

Call $sample() on a StanModel to run MCMC. The result is a StanMCMC object with methods for extracting draws and diagnostics.

fit <- mod$sample(
  data = data,
  iter_warmup = 20,
  iter_sampling = 20,
  chains = 2,
  seed = 123,
  show_messages = FALSE
)

posterior::variables(fit$draws())
#>  [1] "lp__"        "theta"       "log_lik[1]"  "log_lik[2]"  "log_lik[3]" 
#>  [6] "log_lik[4]"  "log_lik[5]"  "log_lik[6]"  "log_lik[7]"  "log_lik[8]" 
#> [11] "log_lik[9]"  "log_lik[10]"

Extract a summary via $summary():

fit$summary()
#> # A tibble: 12 × 10
#>    variable      mean median    sd   mad     q5    q95  rhat ess_bulk ess_tail
#>    <chr>        <dbl>  <dbl> <dbl> <dbl>  <dbl>  <dbl> <dbl>    <dbl>    <dbl>
#>  1 lp__        -8.73  -8.47  0.612 0.210 -9.99  -8.32   1.24     10.6     22.4
#>  2 theta        0.500  0.485 0.126 0.101  0.316  0.700  1.20     21.1     21.4
#>  3 log_lik[1]  -0.727 -0.723 0.282 0.218 -1.15  -0.357  1.16     21.1     21.4
#>  4 log_lik[2]  -0.728 -0.664 0.268 0.197 -1.20  -0.380  1.20     21.1     21.4
#>  5 log_lik[3]  -0.727 -0.723 0.282 0.218 -1.15  -0.357  1.16     21.1     21.4
#>  6 log_lik[4]  -0.727 -0.723 0.282 0.218 -1.15  -0.357  1.16     21.1     21.4
#>  7 log_lik[5]  -0.728 -0.664 0.268 0.197 -1.20  -0.380  1.20     21.1     21.4
#>  8 log_lik[6]  -0.727 -0.723 0.282 0.218 -1.15  -0.357  1.16     21.1     21.4
#>  9 log_lik[7]  -0.728 -0.664 0.268 0.197 -1.20  -0.380  1.20     21.1     21.4
#> 10 log_lik[8]  -0.728 -0.664 0.268 0.197 -1.20  -0.380  1.20     21.1     21.4
#> 11 log_lik[9]  -0.727 -0.723 0.282 0.218 -1.15  -0.357  1.16     21.1     21.4
#> 12 log_lik[10] -0.728 -0.664 0.268 0.197 -1.20  -0.380  1.20     21.1     21.4

Access sampler diagnostics and chain information:

list(
  chains = fit$num_chains(),
  return_codes = fit$return_codes(),
  diagnostics = fit$diagnostic_summary()
)
#> $chains
#> [1] 2
#> 
#> $return_codes
#> [1] 0 0
#> 
#> $diagnostics
#> $diagnostics$num_divergent
#> [1] 0 0
#> 
#> $diagnostics$num_max_treedepth
#> [1] 0 0
#> 
#> $diagnostics$ebfmi
#> [1] 1.2035473 0.9794559

Fixed Param

For models without parameters, use fixed_param = TRUE:

fixed_param_model <- "
data {
  int<lower=0> N;
}
generated quantities {
  array[N] int<lower=0, upper=1> y_rep;
  for (n in 1:N) {
    y_rep[n] = bernoulli_rng(0.5);
  }
}
"

fixed_mod <- stan_model(code = fixed_param_model, model_name = "fixed_param")

fixed_fit <- fixed_mod$sample(
  data = list(N = 5),
  iter_warmup = 0,
  iter_sampling = 10,
  fixed_param = TRUE,
  seed = 123,
  show_messages = FALSE
)

posterior::variables(fixed_fit$draws())
#> [1] "lp__"     "y_rep[1]" "y_rep[2]" "y_rep[3]" "y_rep[4]" "y_rep[5]"

Supplying Inverse Metrics

$sample() accepts precomputed inverse metric values through inv_metric.

For a diagonal metric, pass a numeric vector with one value per unconstrained parameter:

diag_fit <- mod$sample(
  data = data,
  metric = "diag_e",
  inv_metric = c(1.0),
  iter_warmup = 20,
  iter_sampling = 20,
  seed = 123,
  show_messages = FALSE
)

posterior::variables(diag_fit$draws())
#>  [1] "lp__"        "theta"       "log_lik[1]"  "log_lik[2]"  "log_lik[3]" 
#>  [6] "log_lik[4]"  "log_lik[5]"  "log_lik[6]"  "log_lik[7]"  "log_lik[8]" 
#> [11] "log_lik[9]"  "log_lik[10]"

For a dense metric, pass a square matrix. To provide one metric per chain, pass a list whose length equals chains:

dense_fit <- mod$sample(
  data = data,
  metric = "dense_e",
  inv_metric = list(matrix(1.0, nrow = 1, ncol = 1)),
  chains = 1,
  iter_warmup = 20,
  iter_sampling = 20,
  seed = 123,
  show_messages = FALSE
)

posterior::variables(dense_fit$draws())
#>  [1] "lp__"        "theta"       "log_lik[1]"  "log_lik[2]"  "log_lik[3]" 
#>  [6] "log_lik[4]"  "log_lik[5]"  "log_lik[6]"  "log_lik[7]"  "log_lik[8]" 
#> [11] "log_lik[9]"  "log_lik[10]"

WALNUTS Engine

engine = "walnuts" samples with WALNUTS instead of Stan’s NUTS – a locally-adaptive step size within each trajectory, useful for posteriors with varying curvature. It only supports metric = "diag_e" and adapt_engaged = TRUE.

walnuts_fit <- mod$sample(
  data = data,
  engine = "walnuts",
  iter_warmup = 20,
  iter_sampling = 20,
  chains = 2,
  seed = 123,
  show_messages = FALSE
)

walnuts_fit$summary()
#> # A tibble: 12 × 10
#>    variable      mean median    sd   mad      q5    q95  rhat ess_bulk ess_tail
#>    <chr>        <dbl>  <dbl> <dbl> <dbl>   <dbl>  <dbl> <dbl>    <dbl>    <dbl>
#>  1 lp__        -8.79  -8.55  0.615 0.315 -10.1   -8.32   1.10    18.5      49.6
#>  2 theta        0.525  0.536 0.133 0.119   0.317  0.753  1.30     7.70     56.9
#>  3 log_lik[1]  -0.677 -0.623 0.268 0.207  -1.15  -0.284  1.30     7.70     56.9
#>  4 log_lik[2]  -0.788 -0.769 0.307 0.265  -1.40  -0.382  1.30     7.70     56.9
#>  5 log_lik[3]  -0.677 -0.623 0.268 0.207  -1.15  -0.284  1.30     7.70     56.9
#>  6 log_lik[4]  -0.677 -0.623 0.268 0.207  -1.15  -0.284  1.30     7.70     56.9
#>  7 log_lik[5]  -0.788 -0.769 0.307 0.265  -1.40  -0.382  1.30     7.70     56.9
#>  8 log_lik[6]  -0.677 -0.623 0.268 0.207  -1.15  -0.284  1.30     7.70     56.9
#>  9 log_lik[7]  -0.788 -0.769 0.307 0.265  -1.40  -0.382  1.30     7.70     56.9
#> 10 log_lik[8]  -0.788 -0.769 0.307 0.265  -1.40  -0.382  1.30     7.70     56.9
#> 11 log_lik[9]  -0.677 -0.623 0.268 0.207  -1.15  -0.284  1.30     7.70     56.9
#> 12 log_lik[10] -0.788 -0.769 0.307 0.265  -1.40  -0.382  1.30     7.70     56.9

Pathfinder

Call $pathfinder() for an approximate posterior based on L-BFGS paths.

pf <- mod$pathfinder(
  data = data,
  max_lbfgs_iters = 100,
  num_paths = 1,
  draws = 50,
  seed = 123,
  show_messages = FALSE
)

posterior::variables(pf$draws())
#>  [1] "lp_approx__" "lp__"        "path__"      "theta"       "log_lik[1]" 
#>  [6] "log_lik[2]"  "log_lik[3]"  "log_lik[4]"  "log_lik[5]"  "log_lik[6]" 
#> [11] "log_lik[7]"  "log_lik[8]"  "log_lik[9]"  "log_lik[10]"
pf$summary()
#> # A tibble: 14 × 10
#>    variable      mean median    sd   mad      q5    q95   rhat ess_bulk ess_tail
#>    <chr>        <dbl>  <dbl> <dbl> <dbl>   <dbl>  <dbl>  <dbl>    <dbl>    <dbl>
#>  1 lp_approx__ -0.931 -0.653 0.735 0.295  -2.34  -0.439  1.00     1085.     979.
#>  2 lp__        -8.86  -8.57  0.770 0.340 -10.4   -8.32   1.00     1095.     941.
#>  3 path__       1      1     0     0       1      1     NA          NA       NA 
#>  4 theta        0.510  0.515 0.141 0.152   0.275  0.738  0.999     953.     906.
#>  5 log_lik[1]  -0.716 -0.663 0.306 0.287  -1.29  -0.303  0.999     953.     906.
#>  6 log_lik[2]  -0.761 -0.724 0.320 0.306  -1.34  -0.322  0.999     953.     906.
#>  7 log_lik[3]  -0.716 -0.663 0.306 0.287  -1.29  -0.303  0.999     953.     906.
#>  8 log_lik[4]  -0.716 -0.663 0.306 0.287  -1.29  -0.303  0.999     953.     906.
#>  9 log_lik[5]  -0.761 -0.724 0.320 0.306  -1.34  -0.322  0.999     953.     906.
#> 10 log_lik[6]  -0.716 -0.663 0.306 0.287  -1.29  -0.303  0.999     953.     906.
#> 11 log_lik[7]  -0.761 -0.724 0.320 0.306  -1.34  -0.322  0.999     953.     906.
#> 12 log_lik[8]  -0.761 -0.724 0.320 0.306  -1.34  -0.322  0.999     953.     906.
#> 13 log_lik[9]  -0.716 -0.663 0.306 0.287  -1.29  -0.303  0.999     953.     906.
#> 14 log_lik[10] -0.761 -0.724 0.320 0.306  -1.34  -0.322  0.999     953.     906.

Variational Inference

Call $variational() for ADVI with either a mean-field or full-rank Gaussian family.

vb <- mod$variational(
  data = data,
  algorithm = "meanfield",
  iter = 1000,
  draws = 50,
  seed = 123,
  show_messages = FALSE
)

posterior::variables(vb$draws())
#>  [1] "lp_approx__" "theta"       "log_lik[1]"  "log_lik[2]"  "log_lik[3]" 
#>  [6] "log_lik[4]"  "log_lik[5]"  "log_lik[6]"  "log_lik[7]"  "log_lik[8]" 
#> [11] "log_lik[9]"  "log_lik[10]"
vb$summary()
#> # A tibble: 12 × 10
#>    variable      mean median    sd   mad     q5      q95  rhat ess_bulk ess_tail
#>    <chr>        <dbl>  <dbl> <dbl> <dbl>  <dbl>    <dbl> <dbl>    <dbl>    <dbl>
#>  1 lp_approx__ -0.657 -0.181 1.04  0.253 -3.14  -0.00320 0.999     38.4     20.6
#>  2 theta        0.501  0.503 0.144 0.134  0.212  0.712   1.01      84.7     46.0
#>  3 log_lik[1]  -0.743 -0.687 0.352 0.250 -1.55  -0.339   1.01      84.7     46.0
#>  4 log_lik[2]  -0.737 -0.700 0.294 0.259 -1.25  -0.239   1.01      84.7     46.0
#>  5 log_lik[3]  -0.743 -0.687 0.352 0.250 -1.55  -0.339   1.01      84.7     46.0
#>  6 log_lik[4]  -0.743 -0.687 0.352 0.250 -1.55  -0.339   1.01      84.7     46.0
#>  7 log_lik[5]  -0.737 -0.700 0.294 0.259 -1.25  -0.239   1.01      84.7     46.0
#>  8 log_lik[6]  -0.743 -0.687 0.352 0.250 -1.55  -0.339   1.01      84.7     46.0
#>  9 log_lik[7]  -0.737 -0.700 0.294 0.259 -1.25  -0.239   1.01      84.7     46.0
#> 10 log_lik[8]  -0.737 -0.700 0.294 0.259 -1.25  -0.239   1.01      84.7     46.0
#> 11 log_lik[9]  -0.743 -0.687 0.352 0.250 -1.55  -0.339   1.01      84.7     46.0
#> 12 log_lik[10] -0.737 -0.700 0.294 0.259 -1.25  -0.239   1.01      84.7     46.0

Optimization

Call $optimize() to find a posterior mode or maximum likelihood estimate.

opt <- mod$optimize(
  data = data,
  algorithm = "lbfgs",
  seed = 123,
  show_messages = FALSE
)

opt$mle()
#>       theta  log_lik[1]  log_lik[2]  log_lik[3]  log_lik[4]  log_lik[5] 
#>   0.5000001  -0.6931471  -0.6931473  -0.6931471  -0.6931471  -0.6931473 
#>  log_lik[6]  log_lik[7]  log_lik[8]  log_lik[9] log_lik[10] 
#>  -0.6931471  -0.6931473  -0.6931473  -0.6931471  -0.6931473
opt$summary()
#>       variable   estimate
#> 1        theta  0.5000001
#> 2   log_lik[1] -0.6931471
#> 3   log_lik[2] -0.6931473
#> 4   log_lik[3] -0.6931471
#> 5   log_lik[4] -0.6931471
#> 6   log_lik[5] -0.6931473
#> 7   log_lik[6] -0.6931471
#> 8   log_lik[7] -0.6931473
#> 9   log_lik[8] -0.6931473
#> 10  log_lik[9] -0.6931471
#> 11 log_lik[10] -0.6931473

Laplace Approximation

Call $laplace() to draw from a Gaussian approximation around a mode. Pass a StanMLE object or let $laplace() run optimization first.

lap <- mod$laplace(
  data = data,
  mode = opt,
  draws = 20,
  seed = 123,
  show_messages = FALSE
)

posterior::variables(lap$draws())
#>  [1] "lp__"        "lp_approx__" "theta"       "log_lik[1]"  "log_lik[2]" 
#>  [6] "log_lik[3]"  "log_lik[4]"  "log_lik[5]"  "log_lik[6]"  "log_lik[7]" 
#> [11] "log_lik[8]"  "log_lik[9]"  "log_lik[10]"
lap$summary()
#> Warning: The ESS has been capped to avoid unstable estimates.
#> Warning: The ESS has been capped to avoid unstable estimates.
#> Warning: The ESS has been capped to avoid unstable estimates.
#> Warning: The ESS has been capped to avoid unstable estimates.
#> Warning: The ESS has been capped to avoid unstable estimates.
#> Warning: The ESS has been capped to avoid unstable estimates.
#> Warning: The ESS has been capped to avoid unstable estimates.
#> Warning: The ESS has been capped to avoid unstable estimates.
#> Warning: The ESS has been capped to avoid unstable estimates.
#> Warning: The ESS has been capped to avoid unstable estimates.
#> Warning: The ESS has been capped to avoid unstable estimates.
#> # A tibble: 13 × 10
#>    variable      mean median    sd   mad      q5     q95  rhat ess_bulk ess_tail
#>    <chr>        <dbl>  <dbl> <dbl> <dbl>   <dbl>   <dbl> <dbl>    <dbl>    <dbl>
#>  1 lp__        -9.16  -8.78  0.928 0.645 -10.7   -8.33   1.13      19.0     25.6
#>  2 lp_approx__ -0.586 -0.433 0.764 0.597  -1.34  -0.0106 0.985     20.5     20.4
#>  3 theta        0.596  0.609 0.149 0.181   0.386  0.785  0.954     26.0     25.6
#>  4 log_lik[1]  -0.552 -0.497 0.283 0.323  -0.956 -0.242  0.953     26.0     25.6
#>  5 log_lik[2]  -0.974 -0.939 0.384 0.404  -1.54  -0.488  0.972     26.0     25.6
#>  6 log_lik[3]  -0.552 -0.497 0.283 0.323  -0.956 -0.242  0.953     26.0     25.6
#>  7 log_lik[4]  -0.552 -0.497 0.283 0.323  -0.956 -0.242  0.953     26.0     25.6
#>  8 log_lik[5]  -0.974 -0.939 0.384 0.404  -1.54  -0.488  0.972     26.0     25.6
#>  9 log_lik[6]  -0.552 -0.497 0.283 0.323  -0.956 -0.242  0.953     26.0     25.6
#> 10 log_lik[7]  -0.974 -0.939 0.384 0.404  -1.54  -0.488  0.972     26.0     25.6
#> 11 log_lik[8]  -0.974 -0.939 0.384 0.404  -1.54  -0.488  0.972     26.0     25.6
#> 12 log_lik[9]  -0.552 -0.497 0.283 0.323  -0.956 -0.242  0.953     26.0     25.6
#> 13 log_lik[10] -0.974 -0.939 0.384 0.404  -1.54  -0.488  0.972     26.0     25.6

Generated Quantities

Call $generate_quantities() with a StanFit object or draws matrix to evaluate the generated quantities block.

gq <- mod$generate_quantities(
  fitted_params = fit,
  data = data,
  seed = 123,
  show_messages = FALSE
)

posterior::variables(gq$draws())
#>  [1] "log_lik[1]"  "log_lik[2]"  "log_lik[3]"  "log_lik[4]"  "log_lik[5]" 
#>  [6] "log_lik[6]"  "log_lik[7]"  "log_lik[8]"  "log_lik[9]"  "log_lik[10]"
gq$summary()
#> # A tibble: 10 × 10
#>    variable      mean median    sd   mad    q5    q95  rhat ess_bulk ess_tail
#>    <chr>        <dbl>  <dbl> <dbl> <dbl> <dbl>  <dbl> <dbl>    <dbl>    <dbl>
#>  1 log_lik[1]  -0.727 -0.723 0.282 0.218 -1.15 -0.357  1.09     16.0     21.1
#>  2 log_lik[2]  -0.728 -0.664 0.268 0.197 -1.20 -0.380  1.09     16.0     21.1
#>  3 log_lik[3]  -0.727 -0.723 0.282 0.218 -1.15 -0.357  1.09     16.0     21.1
#>  4 log_lik[4]  -0.727 -0.723 0.282 0.218 -1.15 -0.357  1.09     16.0     21.1
#>  5 log_lik[5]  -0.728 -0.664 0.268 0.197 -1.20 -0.380  1.09     16.0     21.1
#>  6 log_lik[6]  -0.727 -0.723 0.282 0.218 -1.15 -0.357  1.09     16.0     21.1
#>  7 log_lik[7]  -0.728 -0.664 0.268 0.197 -1.20 -0.380  1.09     16.0     21.1
#>  8 log_lik[8]  -0.728 -0.664 0.268 0.197 -1.20 -0.380  1.09     16.0     21.1
#>  9 log_lik[9]  -0.727 -0.723 0.282 0.218 -1.15 -0.357  1.09     16.0     21.1
#> 10 log_lik[10] -0.728 -0.664 0.268 0.197 -1.20 -0.380  1.09     16.0     21.1

Gradient Check

Call $diagnose() to compare autodiff gradients with finite-difference approximations.

diag <- mod$diagnose(
  data = data,
  seed = 123
)
#> TEST GRADIENT MODE
#> 
#>  Log probability=-12.776
#> 
#>  param idx           value           model     finite diff           error
#>          0         1.83245        -4.34465        -4.34465      2.2707e-09
#> [stanr] All gradient tests passed.

list(
  num_failed = diag$num_failed(),
  lp = diag$lp(),
  gradients = diag$gradients()
)
#> $num_failed
#> [1] 0
#> 
#> $lp
#> [1] -12.77597
#> 
#> $gradients
#>   param_idx    value     model finite_diff        error
#> 1         0 1.832454 -4.344645   -4.344645 2.270696e-09

Tuples and Complex Numbers

Stan tuple(...) values map to/from unnamed R lists (one element per slot; arrays of tuples become lists of such lists), and complex / complex_vector / complex_row_vector / complex_matrix map to/from R’s native complex type – for data =, init =, draws, and exposed functions alike.

tuple_complex_code <- "
functions {
  tuple(real, vector) split_stat(vector x) {
    return (mean(x), head(x, 2));
  }
  complex_vector rotate(complex_vector z, complex phase) {
    return z * phase;
  }
}
"

tc_mod <- stan_model(
  code = tuple_complex_code,
  model_name = "tuple_complex",
  compile = FALSE
)
tc_mod$expose_stan_functions()

# tuple(real, vector) comes back as an unnamed list: the scalar slot, then
# the vector slot.
tc_mod$functions$split_stat(c(1, 2, 3, 4))
#> [[1]]
#> [1] 2.5
#> 
#> [[2]]
#> [1] 1 2

# complex_vector arguments/returns round-trip as native R complex vectors.
tc_mod$functions$rotate(c(1 + 2i, 3 - 1i), 1i)
#> [1] -2+1i  1+3i

Stan Compilation Helpers

stanc() compiles Stan code to C++ and returns the generated C++ source as a single string.

cpp_code <- stanc(bernoulli_model)
substr(cpp_code, 1, 80)
#> [1] "// Code generated by stanc 4d256d1\n#include <stan/model/model_header.hpp>\nnamesp"

stan_model() is the higher-level entry point: it calls stanc(), compiles the generated C++, and returns a StanModel object used by all inference methods.

Name		Name	Last commit message	Last commit date
.github		.github
R		R
inst		inst
man-roxygen		man-roxygen
man		man
src		src
tests		tests
tools		tools
.Rbuildignore		.Rbuildignore
.gitignore		.gitignore
DESCRIPTION		DESCRIPTION
LICENSE		LICENSE
LICENSE.md		LICENSE.md
NAMESPACE		NAMESPACE
README.Rmd		README.Rmd
README.md		README.md

Navigation Menu

Repository files navigation

stanr

Inference Methods

Fit Object Methods

Installation

Compile a Model

Sampling

Fixed Param

Supplying Inverse Metrics

WALNUTS Engine

Pathfinder

Variational Inference

Optimization

Laplace Approximation

Generated Quantities

Gradient Check

Tuples and Complex Numbers

Stan Compilation Helpers

About

Releases

Packages

Contributors

Languages

Latest commit

History

Folders and files

Repository files navigation

stanr

Inference Methods

Fit Object Methods

Installation

Compile a Model

Sampling

Fixed Param

Supplying Inverse Metrics

WALNUTS Engine

Pathfinder

Variational Inference

Optimization

Laplace Approximation

Generated Quantities

Gradient Check

Tuples and Complex Numbers

Stan Compilation Helpers

About

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages