CmdStanR

https://github.com/stan-dev/cmdstanr
Admin
CmdStanR is a lightweight R interface to Stan that enables Bayesian statistical modeling by calling...
Tokens:9,195
Snippets:75
Trust Score:9.4
Update:6 months ago
Show doc for...
Tokens:
Raw
### Install and Configure CmdStan

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

Functions to download, install, and verify the CmdStan toolchain. These utilities ensure the environment is correctly set up for compiling Stan models.

```r
library(cmdstanr)
install_cmdstan(cores = 4)
install_cmdstan(version = "2.35.0", cores = 4)
check_cmdstan_toolchain()
set_cmdstan_path("/path/to/cmdstan-2.35.0")
```

--------------------------------

### Fit Example CmdStan Models

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

Fits built-in example Stan models provided with CmdStanR for demonstration, testing, and examples. Supports different inference methods (optimize, variational, pathfinder) and printing Stan code.

```r
library(cmdstanr)

# Logistic regression example
fit_logistic <- cmdstanr_example("logistic", chains = 4)
fit_logistic$summary()

# Eight schools hierarchical model (will show divergences)
fit_schools <- cmdstanr_example("schools", quiet = FALSE)

# Non-centered parameterization (fixes divergences)
fit_schools_ncp <- cmdstanr_example("schools_ncp", chains = 4)

# Use different methods
fit_optim <- cmdstanr_example("logistic", method = "optimize")
fit_vb <- cmdstanr_example("logistic", method = "variational")
fit_pf <- cmdstanr_example("logistic", method = "pathfinder")

# Print example Stan code
print_example_program("logistic")
print_example_program("schools")
```

--------------------------------

### Install CmdStanR R Package

Source: https://github.com/stan-dev/cmdstanr/blob/master/README.md

Installs the latest release of the cmdstanr R package from a specified repository. It's recommended to run this in a fresh R session or restart the current one.

```R
install.packages("cmdstanr", repos = c('https://stan-dev.r-universe.dev', getOption("repos")))
```

--------------------------------

### cmdstanr_example

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

Fits built-in example models for demonstration and testing purposes, supporting various inference methods.

```APIDOC
## cmdstanr_example

### Description
Fits pre-defined example Stan models included with the `cmdstanr` package. Useful for testing, demonstration, and learning. Supports different inference methods like MCMC, optimization, variational Bayes, and pathfinder.

### Method
Not applicable (R function)

### Endpoint
Not applicable (R function)

### Parameters
#### Path Parameters
None

#### Query Parameters
None

#### Request Body
None

### Request Example
```r
library(cmdstanr)

# Logistic regression example
fit_logistic <- cmdstanr_example("logistic", chains = 4)
fit_logistic$summary()

# Eight schools hierarchical model (will show divergences)
fit_schools <- cmdstanr_example("schools", quiet = FALSE)

# Non-centered parameterization (fixes divergences)
fit_schools_ncp <- cmdstanr_example("schools_ncp", chains = 4)

# Use different methods
fit_optim <- cmdstanr_example("logistic", method = "optimize")
fit_vb <- cmdstanr_example("logistic", method = "variational")
fit_pf <- cmdstanr_example("logistic", method = "pathfinder")

# Print example Stan code
print_example_program("logistic")
print_example_program("schools")
```

### Response
#### Success Response (200)
Returns a `CmdStanFit` object representing the fitted model.

#### Response Example
```r
# CmdStanFit object (conceptual)
# <CmdStanFit>
```
```

--------------------------------

### Install CmdStanR Development Version from GitHub

Source: https://github.com/stan-dev/cmdstanr/blob/master/README.md

Installs the latest development version of the cmdstanr R package directly from GitHub using the remotes package.

```R
# install.packages("remotes")
remotes::install_github("stan-dev/cmdstanr")
```

--------------------------------

### Install CmdStan via R

Source: https://github.com/stan-dev/cmdstanr/blob/master/vignettes/children/comparison-with-rstan.md

This command downloads and installs the latest version of CmdStan, which is required for the CmdStanR package to compile and run Stan models.

```R
cmdstanr::install_cmdstan()
```

--------------------------------

### Get Diagnostic Summary with $diagnostic_summary() in CmdStanR

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

Illustrates how to retrieve diagnostic information from a fitted CmdStanR model using $diagnostic_summary(). This includes counts of divergent transitions, maximum treedepth hits, and E-BFMI values for each chain.

```r
library(cmdstanr)

# Fit a model that may have issues
fit <- cmdstanr_example("schools")

# Get diagnostic summary
diagnostics <- fit$diagnostic_summary()
print(diagnostics)

# Total divergences
sum(diagnostics$num_divergent)

# Check for low E-BFMI
any(diagnostics$ebfmi < 0.3)
```

--------------------------------

### CmdStan Model Information Methods

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

Retrieves metadata and file information from a CmdStan fit object. Includes methods to get output files, model metadata, timing information, number of chains, return codes, and raw output for specific chains.

```r
library(cmdstanr)

fit <- cmdstanr_example("logistic", chains = 4)

# Get output files
fit$output_files()

# Get model metadata
fit$metadata()

# Get timing information
fit$time()
# $total
# [1] 0.5
# $chains
#   chain_id warmup sampling total
# 1        1   0.05     0.07  0.12
# 2        2   0.06     0.08  0.14
# ...

# Number of chains
fit$num_chains()

# Return codes
fit$return_codes()

# Get raw output
fit$output(1)  # Output from chain 1
```

--------------------------------

### GET /summary

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

Computes summary statistics for posterior draws.

```APIDOC
## GET /summary

### Description
Calculates summary statistics such as mean, sd, and quantiles for model parameters.

### Method
GET

### Endpoint
fit$summary()

### Parameters
#### Query Parameters
- **variables** (vector) - Optional - Variables to summarize.

### Response
#### Success Response (200)
- **summary** (data.frame) - Table containing summary statistics.
```

--------------------------------

### Write Data to JSON for CmdStan with write_stan_json

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

Provides an example of using the write_stan_json function to prepare R data structures (lists, matrices, vectors) into the JSON format required by CmdStan. It handles basic types and nested structures, including arrays of vectors.

```r
library(cmdstanr)

# Prepare data
x <- matrix(rnorm(10), 5, 2)
y <- rpois(nrow(x), lambda = 10)
z <- c(TRUE, FALSE)
data <- list(N = nrow(x), K = ncol(x), x = x, y = y, z = z)

# Write to JSON file
file <- tempfile(fileext = ".json")
write_stan_json(data, file)

# Check contents
cat(readLines(file), sep = "\n")

# Array of vectors example (for array[K] vector[J] declaration)
data <- list(x = list(1:3, 4:6))
write_stan_json(data, file)
```

--------------------------------

### GET /draws

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

Extracts posterior draws from a fitted model object in various formats.

```APIDOC
## GET /draws

### Description
Retrieves posterior draws from the model fit in formats like array, data frame, or matrix.

### Method
GET

### Endpoint
fit$draws()

### Parameters
#### Query Parameters
- **format** (string) - Optional - Output format (e.g., 'df', 'matrix').
- **variables** (vector) - Optional - Specific variables to extract.
- **inc_warmup** (boolean) - Optional - Include warmup iterations.

### Response
#### Success Response (200)
- **draws** (array/matrix/df) - The extracted posterior samples.
```

--------------------------------

### Run MCMC Sampling with CmdStanR

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

Demonstrates how to execute MCMC sampling using the $sample() method, including basic configuration, summary statistics extraction, and diagnostic checks.

```r
fit <- mod$sample(data = data_list, seed = 123, chains = 4, parallel_chains = 4, iter_warmup = 1000, iter_sampling = 1000, refresh = 500)
fit$summary()
draws_arr <- fit$draws()
draws_df <- fit$draws(format = "df")
fit$sampler_diagnostics()
```

--------------------------------

### Creating CmdStanFit Object

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

Creates a CmdStanFit object from CSV files, enabling further analysis and summarization.

```APIDOC
## as_cmdstan_fit

### Description
Creates a CmdStanFit object from a list of CSV files, which can then be used to generate summaries and diagnostics.

### Method
Not applicable (R function)

### Endpoint
Not applicable (R function)

### Parameters
#### Path Parameters
None

#### Query Parameters
None

#### Request Body
None

### Request Example
```r
# Create fit object from CSV files
fit2 <- as_cmdstan_fit(csv_files)
fit2$summary()
```

### Response
#### Success Response (200)
Returns a CmdStanFit object.

#### Response Example
```r
# CmdStanFit object (conceptual)
# <CmdStanFit>
```
```

--------------------------------

### POST /sample

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

Executes MCMC sampling on a compiled Stan model to generate posterior draws.

```APIDOC
## POST $sample()

### Description
Runs MCMC sampling to estimate the posterior distribution of model parameters.

### Method
POST

### Parameters
#### Request Body
- **data** (list/string) - Required - Data list or path to a JSON data file.
- **seed** (integer) - Optional - Random number seed.
- **chains** (integer) - Optional - Number of MCMC chains.
- **iter_warmup** (integer) - Optional - Number of warmup iterations.
- **iter_sampling** (integer) - Optional - Number of sampling iterations.
- **adapt_delta** (double) - Optional - Target acceptance rate for NUTS.

### Request Example
{
  "data": {"N": 100, "y": [0, 1, 0]},
  "chains": 4,
  "iter_sampling": 1000
}

### Response
#### Success Response (200)
- **fit** (object) - Returns a CmdStanMCMC object containing draws and diagnostic information.
```

--------------------------------

### Model Information Methods

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

Provides access to model metadata, output files, and timing information from a fitted CmdStan model.

```APIDOC
## Model Information Methods

### Description
Retrieve various pieces of information about a fitted CmdStan model, including the paths to output files, model metadata, timing information, return codes, and raw output for individual chains.

### Method
Not applicable (R methods on CmdStanFit object)

### Endpoint
Not applicable

### Parameters
#### Path Parameters
None

#### Query Parameters
None

#### Request Body
None

### Request Example
```r
library(cmdstanr)

fit <- cmdstanr_example("logistic", chains = 4)

# Get output files
fit$output_files()

# Get model metadata
fit$metadata()

# Get timing information
fit$time()
# $total
# [1] 0.5
# $chains
#   chain_id warmup sampling total
# 1        1   0.05     0.07  0.12
# 2        2   0.06     0.08  0.14
# ...

# Number of chains
fit$num_chains()

# Return codes
fit$return_codes()

# Get raw output
fit$output(1)  # Output from chain 1
```

### Response
#### Success Response (200)
Returns information such as character vectors of file paths, a list of metadata, a list of timing information, an integer for the number of chains, a list of return codes, or character strings for raw output.

#### Response Example
```r
# Example output for output_files()
# ["path/to/output-chain1.csv", "path/to/output-chain2.csv"]

# Example output for time()
# list(total = 0.5, chains = data.frame(...))
```
```

--------------------------------

### Run Pathfinder Sampling in CmdStanR

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

Demonstrates how to use the pathfinder function for single-path and multi-path (recommended) approximate posterior sampling. It takes data, a seed, and the number of draws as input. Multi-path includes options for num_paths and psis_resample for improved importance sampling.

```r
fit_pf <- mod$pathfinder(
  data = data_list,
  seed = 123,
  draws = 4000
)

fit_pf$summary()

fit_pf_multi <- mod$pathfinder(
  data = data_list,
  seed = 123,
  draws = 4000,
  num_paths = 4,
  psis_resample = TRUE  # PSIS resampling for importance sampling
)
```

--------------------------------

### CmdStan Model Diagnostics Methods

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

Provides access to CmdStan's diagnostic utilities through R methods. Includes running CmdStan's 'diagnose' and 'stansummary' utilities, retrieving sampler diagnostics as draws or data frames, and summarizing diagnostic information.

```r
library(cmdstanr)

fit <- cmdstanr_example("schools", chains = 4)

# Run CmdStan's diagnose utility
fit$cmdstan_diagnose()

# Run CmdStan's stansummary utility
fit$cmdstan_summary()

# Get sampler diagnostics as draws
sampler_diag <- fit$sampler_diagnostics()
str(sampler_diag)
# Variables: accept_stat__, stepsize__, treedepth__,
#            n_leapfrog__, divergent__, energy__

# As data frame
sampler_diag_df <- fit$sampler_diagnostics(format = "df")

# Diagnostic summary with warnings
diagnostics <- fit$diagnostic_summary()
```

--------------------------------

### Configure Advanced MCMC Sampling Options

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

Covers advanced sampling parameters such as multi-threading, thinning, adaptation control, and custom initialization functions for complex models.

```r
fit <- mod$sample(data = data_list, seed = 123, chains = 4, parallel_chains = 4, threads_per_chain = 2, iter_warmup = 2000, iter_sampling = 2000, thin = 2, adapt_delta = 0.95, max_treedepth = 12, init = 0.5, save_warmup = TRUE, save_latent_dynamics = TRUE, output_dir = "output")
init_fun <- function() list(theta = runif(1, 0.1, 0.9))
fit <- mod$sample(data = data_list, init = init_fun, chains = 4)
```

--------------------------------

### Generate Quantities of Interest with CmdStanR

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

Shows how to use the $generate_quantities method to compute derived quantities, posterior predictions, or log-likelihoods from existing parameter draws without re-running MCMC. It requires a fitted model object and the data used for fitting.

```r
library(cmdstanr)

# Original model
mod <- cmdstan_model(write_stan_file("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);
}"))

data_list <- list(N = 10, y = c(0, 1, 0, 0, 0, 0, 0, 0, 0, 1))
fit <- mod$sample(data = data_list, chains = 4, seed = 123)

# Model with generated quantities
mod_gq <- cmdstan_model(write_stan_file("data {
  int<lower=0> N;
  array[N] int<lower=0,upper=1> y;
}
parameters {
  real<lower=0,upper=1> theta;
}
generated quantities {
  array[N] int y_rep;
  real log_lik = 0;
  for (n in 1:N) {
    y_rep[n] = bernoulli_rng(theta);
    log_lik += bernoulli_lpmf(y[n] | theta);
  }
}"))

# Generate quantities using fitted parameters
fit_gq <- mod_gq$generate_quantities(
  fitted_params = fit,
  data = data_list,
  seed = 123
)

# Extract generated quantities
y_rep <- fit_gq$draws("y_rep")
log_lik <- fit_gq$draws("log_lik")
```

--------------------------------

### Model Diagnostics Methods

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

Provides access to diagnostic utilities for CmdStan models, including `diagnose`, `summary`, and sampler diagnostics.

```APIDOC
## Model Diagnostics Methods

### Description
Access and utilize diagnostic utilities provided by CmdStan through the `CmdStanFit` object. This includes running CmdStan's `diagnose` and `stansummary` utilities, and extracting detailed sampler diagnostics.

### Method
Not applicable (R methods on CmdStanFit object)

### Endpoint
Not applicable

### Parameters
#### Path Parameters
None

#### Query Parameters
None

#### Request Body
None

### Request Example
```r
library(cmdstanr)

fit <- cmdstanr_example("schools", chains = 4)

# Run CmdStan's diagnose utility
fit$cmdstan_diagnose()

# Run CmdStan's stansummary utility
fit$cmdstan_summary()

# Get sampler diagnostics as draws
sampler_diag <- fit$sampler_diagnostics()
str(sampler_diag)
# Variables: accept_stat__, stepsize__, treedepth__,
#            n_leapfrog__, divergent__, energy__

# As data frame
sampler_diag_df <- fit$sampler_diagnostics(format = "df")

# Diagnostic summary with warnings
diagnostics <- fit$diagnostic_summary()
```

### Response
#### Success Response (200)
Methods return various outputs depending on the utility called: character strings for `cmdstan_diagnose` and `cmdstan_summary`, posterior draws objects or data frames for `sampler_diagnostics`, and a list for `diagnostic_summary`.

#### Response Example
```r
# Example output for sampler_diagnostics()
# posterior::draws_array(...)

# Example output for diagnostic_summary()
# list(
#   warnings = "...",
#   diagnostics = data.frame(...)
# )
```
```

--------------------------------

### Run MCMC Sampling with $sample

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

Executes Hamiltonian Monte Carlo (NUTS) to draw samples from the posterior distribution. Requires a compiled model object and input data.

```r
library(cmdstanr)
mod <- cmdstan_model("model.stan")
data_list <- list(N = 10, y = c(0, 1, 0, 0, 0, 0, 0, 0, 0, 1))
fit <- mod$sample(data = data_list)
```

--------------------------------

### Save Fitted Model Object with $save_object() in CmdStanR

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

Demonstrates saving a fitted CmdStanR model object to disk for later use with $save_object(). It supports saving in RDS (default) or qs2 format, and shows how to load the object back into R.

```r
library(cmdstanr)

fit <- cmdstanr_example("logistic", chains = 4)

# Save as RDS (default)
fit$save_object(file = "fit.RDS")

# Load later
fit_loaded <- readRDS("fit.RDS")
fit_loaded$summary()

# Save as qs2 format (faster for large objects, requires qs2 package)
fit$save_object(file = "fit.qs2", format = "qs2")

# Load qs2 file
fit_loaded <- qs2::qs_read("fit.qs2")
```

--------------------------------

### Compute Summary Statistics with $summary() in CmdStanR

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

Explains how to compute summary statistics for posterior draws using the $summary() method, leveraging the posterior package. It allows for default summaries, summaries of specific variables, custom summary functions, and probability calculations.

```r
library(cmdstanr)
library(posterior)

fit <- cmdstanr_example("logistic", chains = 4)

# Default summary
fit$summary()

# Summary for specific variables
fit$summary(variables = c("alpha", "beta"))

# Custom summary functions
fit$summary(
  variables = "alpha",
  mean, sd, median, mad,
  ~quantile(.x, probs = c(0.025, 0.975))
)

# Probability calculations
fit$summary("alpha", pr_positive = ~ mean(. > 0))

# Multiple custom summaries
fit$summary(
  variables = NULL,  # All variables
  posterior::default_summary_measures(),
  extra_quantiles = ~posterior::quantile2(., probs = c(0.025, 0.975))
)
```

--------------------------------

### Run Variational Inference (ADVI and Pathfinder)

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

Performs approximate Bayesian inference using Automatic Differentiation Variational Inference (ADVI) or the Pathfinder algorithm for faster, stable results.

```r
fit_vb <- mod$variational(data = data_list, seed = 123, draws = 4000, iter = 10000, algorithm = "meanfield")
fit_vb_fullrank <- mod$variational(data = data_list, algorithm = "fullrank", seed = 123)
```

--------------------------------

### Perform Optimization with $optimize()

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

Finds the mode of the posterior distribution or maximum likelihood estimate using various algorithms like BFGS or Newton's method.

```r
fit_mle <- mod$optimize(data = data_list, seed = 123)
fit_map <- mod$optimize(data = data_list, seed = 123, jacobian = TRUE)
fit_bfgs <- mod$optimize(data = data_list, algorithm = "bfgs", iter = 10000, seed = 123)
fit_newton <- mod$optimize(data = data_list, algorithm = "newton", seed = 123)
```

--------------------------------

### Write Stan Files with write_stan_file

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

A utility to write Stan code strings to disk. It uses content hashing for deterministic file naming, which is useful for caching models.

```r
library(cmdstanr)
stan_program <- "data { int<lower=0> N; } parameters { real theta; } model { theta ~ normal(0, 1); }"
file <- write_stan_file(stan_program, dir = "models", basename = "my_model")
```

--------------------------------

### Read CmdStan CSV Output with read_cmdstan_csv

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

Demonstrates how to use the read_cmdstan_csv function to parse CmdStan's CSV output files into R data structures. It can also be used to create a CmdStanR fit object from existing CSV files.

```r
library(cmdstanr)

# Run sampling and get CSV files
fit <- cmdstanr_example("logistic", method = "sample", save_warmup = TRUE)
csv_files <- fit$output_files()

# Read CSV files
x <- read_cmdstan_csv(csv_files)
str(x)
```

--------------------------------

### POST /variational

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

Runs Automatic Differentiation Variational Inference (ADVI) for approximate Bayesian inference.

```APIDOC
## POST $variational()

### Description
Approximates the posterior distribution using Variational Inference.

### Method
POST

### Parameters
#### Request Body
- **data** (list) - Required - Input data.
- **algorithm** (string) - Optional - 'meanfield' or 'fullrank'.
- **draws** (integer) - Optional - Number of draws to generate from the approximation.

### Request Example
{
  "data": {"N": 10, "y": [0, 1, 0]},
  "algorithm": "meanfield"
}

### Response
#### Success Response (200)
- **fit** (object) - Returns a CmdStanVB object containing the variational approximation.
```

--------------------------------

### POST /optimize

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

Finds the mode of the posterior distribution (MAP) or maximum likelihood estimate (MLE).

```APIDOC
## POST $optimize()

### Description
Uses optimization algorithms to find point estimates for model parameters.

### Method
POST

### Parameters
#### Request Body
- **data** (list) - Required - Input data.
- **algorithm** (string) - Optional - Optimization algorithm (e.g., 'bfgs', 'newton').
- **jacobian** (boolean) - Optional - Whether to apply Jacobian adjustment.

### Request Example
{
  "data": {"N": 10, "y": [0, 1, 0]},
  "algorithm": "bfgs"
}

### Response
#### Success Response (200)
- **fit** (object) - Returns a CmdStanOptimize object with point estimates.
```

--------------------------------

### Compile Stan Models with cmdstan_model

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

Compiles a Stan program into an executable. It supports custom compiler options and forced recompilation for iterative development.

```r
library(cmdstanr)
mod <- cmdstan_model("model.stan")
mod_threads <- cmdstan_model("model.stan", cpp_options = list(stan_threads = TRUE))
mod$print()
```

--------------------------------

### Reading CSV from Optimization

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

Reads CSV output specifically from optimization runs.

```APIDOC
## read_cmdstan_csv (optimization)

### Description
Reads CSV output generated from an optimization run using CmdStan.

### Method
Not applicable (R function)

### Endpoint
Not applicable (R function)

### Parameters
#### Path Parameters
None

#### Query Parameters
None

#### Request Body
None

### Request Example
```r
# Read CSV from optimization
fit_opt <- cmdstanr_example("logistic", method = "optimize")
x_opt <- read_cmdstan_csv(fit_opt$output_files())
```

### Response
#### Success Response (200)
Returns a data structure containing the optimization results.

#### Response Example
```r
# Example of returned data structure (conceptual)
# list(
#   optimization_results = ...
# )
```
```

--------------------------------

### POST /generate_quantities

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

Generates quantities of interest from existing parameter draws without re-running MCMC.

```APIDOC
## POST /generate_quantities

### Description
Computes derived quantities, posterior predictions, or log-likelihoods using existing fitted parameters.

### Method
POST

### Endpoint
mod_gq$generate_quantities()

### Parameters
#### Request Body
- **fitted_params** (object) - Required - The existing fit object.
- **data** (list) - Required - The data list.
- **seed** (integer) - Optional - Random seed.

### Request Example
{
  "fitted_params": "fit",
  "data": "data_list",
  "seed": 123
}

### Response
#### Success Response (200)
- **fit_gq** (object) - A new fit object containing the generated quantities.
```

--------------------------------

### Extract Posterior Draws using $draws() in CmdStanR

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

Demonstrates how to extract posterior draws from a fitted CmdStanR model object using the $draws() method. It supports various output formats like draws_array, data frame, and matrix, and allows selection of specific variables and inclusion of warmup draws.

```r
library(cmdstanr)
library(posterior)

fit <- cmdstanr_example("logistic", chains = 4)

# Default: draws_array (iterations x chains x variables)
draws_arr <- fit$draws()
str(draws_arr)

# As data frame
draws_df <- fit$draws(format = "df")
head(draws_df)

# As matrix (all chains combined)
draws_mat <- fit$draws(format = "matrix")

# Select specific variables
theta_draws <- fit$draws(variables = "alpha")
beta_draws <- fit$draws(variables = c("beta[1]", "beta[2]"))

# Include warmup draws
all_draws <- fit$draws(inc_warmup = TRUE)
```

--------------------------------

### POST /pathfinder

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

Performs Pathfinder optimization to approximate the posterior distribution, supporting both single-path and multi-path configurations.

```APIDOC
## POST /pathfinder

### Description
Runs the Pathfinder algorithm to approximate the posterior distribution of a Stan model.

### Method
POST

### Endpoint
mod$pathfinder()

### Parameters
#### Request Body
- **data** (list) - Required - The data list for the model.
- **seed** (integer) - Optional - Random seed for reproducibility.
- **draws** (integer) - Optional - Number of draws to return.
- **num_paths** (integer) - Optional - Number of paths for multi-path optimization.
- **psis_resample** (boolean) - Optional - Whether to use PSIS resampling.

### Request Example
{
  "data": "data_list",
  "seed": 123,
  "draws": 4000,
  "num_paths": 4,
  "psis_resample": true
}

### Response
#### Success Response (200)
- **fit_pf** (object) - A fit object containing the Pathfinder results.
```

--------------------------------

### Execute Laplace Approximation

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

Generates draws from a normal approximation centered at the posterior mode, either using a pre-computed mode or running optimization internally.

```r
fit_map <- mod$optimize(data = data_list, jacobian = TRUE, seed = 123)
fit_laplace <- mod$laplace(mode = fit_map, data = data_list, draws = 4000, seed = 123)
fit_laplace_auto <- mod$laplace(data = data_list, draws = 4000, seed = 123, jacobian = TRUE)
```

--------------------------------

### Convert Posterior Draws to CSV

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

Converts posterior draws objects (from the 'posterior' package) into CSV files. These CSV files are formatted for use with CmdStan's generate_quantities utility. Allows custom output directories and basenames.

```r
library(cmdstanr)
library(posterior)

# Get some draws
draws <- posterior::example_draws()

# Convert to CSV files
csv_files <- draws_to_csv(draws)
print(csv_files)

# With custom directory and basename
csv_files <- draws_to_csv(
  draws,
  dir = "output",
  basename = "my_samples"
)
```

--------------------------------

### Manage Compilation Flags with cmdstan_make_local

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

Configures persistent makefile flags for CmdStan compilation. This allows users to enable features like multi-threading or specific compiler optimizations.

```r
library(cmdstanr)
cmdstan_make_local(cpp_options = list(STAN_THREADS = TRUE, STAN_CPP_OPTIMS = TRUE))
rebuild_cmdstan(cores = 4)
```

--------------------------------

### Reading CmdStan CSV Output

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

Functions to read CSV files generated by CmdStan, with options to specify variables and exclude sampler diagnostics.

```APIDOC
## read_cmdstan_csv

### Description
Reads CSV files generated by CmdStan, allowing for selective loading of variables and exclusion of sampler diagnostics.

### Method
Not applicable (R function)

### Endpoint
Not applicable (R function)

### Parameters
#### Path Parameters
None

#### Query Parameters
None

#### Request Body
None

### Request Example
```r
# Read specific variables
x <- read_cmdstan_csv(csv_files, variables = c("alpha", "beta[1]"))

# Read without sampler diagnostics
x <- read_cmdstan_csv(csv_files, sampler_diagnostics = "")
```

### Response
#### Success Response (200)
Returns a data structure containing the read data.

#### Response Example
```r
# Example of returned data structure (conceptual)
# list(
#   variables = list(
#     alpha = ..., 
#     beta = ...
#   ),
#   metadata = ...
# )
```
```

--------------------------------

### draws_to_csv

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

Converts posterior draws objects to CSV files compatible with CmdStan's generate_quantities.

```APIDOC
## draws_to_csv

### Description
Converts posterior draws objects (from the `posterior` package) into CSV files that can be used with CmdStan's `generate_quantities` utility.

### Method
Not applicable (R function)

### Endpoint
Not applicable (R function)

### Parameters
#### Path Parameters
None

#### Query Parameters
None

#### Request Body
None

### Request Example
```r
library(cmdstanr)
library(posterior)

# Get some draws
draws <- posterior::example_draws()

# Convert to CSV files
csv_files <- draws_to_csv(draws)
print(csv_files)

# With custom directory and basename
csv_files <- draws_to_csv(
  draws,
  dir = "output",
  basename = "my_samples"
)
```

### Response
#### Success Response (200)
Returns a character vector of file paths to the generated CSV files.

#### Response Example
```r
# Example output
# ["output/my_samples-001.csv", "output/my_samples-002.csv"]
```
```

--------------------------------

### Read CmdStan CSV Files

Source: https://context7.com/stan-dev/cmdstanr/llms.txt

Reads CSV output files generated by CmdStan. Supports reading specific variables, excluding sampler diagnostics, and creating CmdStan fit objects. Useful for post-processing CmdStan runs.

```r
library(cmdstanr)

# Read only specific variables
x <- read_cmdstan_csv(csv_files, variables = c("alpha", "beta[1]"))

# Read without sampler diagnostics
x <- read_cmdstan_csv(csv_files, sampler_diagnostics = "")

# Create fit object from CSV files
fit2 <- as_cmdstan_fit(csv_files)
fit2$summary()

# Read CSV from optimization
fit_opt <- cmdstanr_example("logistic", method = "optimize")
x_opt <- read_cmdstan_csv(fit_opt$output_files())
```
Tokens:
Raw