Shortest Probability Interval (SPI)

Compute the Shortest Probability Interval (SPI) of posterior distributions. The SPI is a more computationally stable HDI. The implementation is based on the algorithm from the SPIn package.

Usage

spi(x, ...)

# S3 method for class 'numeric'
spi(x, ci = 0.95, verbose = TRUE, ...)

# S3 method for class 'data.frame'
spi(x, ci = 0.95, rvar_col = NULL, verbose = TRUE, ...)

# S3 method for class 'stanreg'
spi(
  x,
  ci = 0.95,
  effects = c("fixed", "random", "all"),
  component = c("location", "all", "conditional", "smooth_terms", "sigma",
    "distributional", "auxiliary"),
  parameters = NULL,
  verbose = TRUE,
  ...
)

# S3 method for class 'brmsfit'
spi(
  x,
  ci = 0.95,
  effects = c("fixed", "random", "all"),
  component = c("conditional", "zi", "zero_inflated", "all"),
  parameters = NULL,
  verbose = TRUE,
  ...
)

# S3 method for class 'get_predicted'
spi(x, ci = 0.95, use_iterations = FALSE, verbose = TRUE, ...)

Arguments

x: Vector representing a posterior distribution, or a data frame of such vectors. Can also be a Bayesian model. bayestestR supports a wide range of models (see, for example, methods("hdi")) and not all of those are documented in the 'Usage' section, because methods for other classes mostly resemble the arguments of the .numeric or .data.framemethods.
...: Currently not used.
ci: Value or vector of probability of the (credible) interval - CI (between 0 and 1) to be estimated. Default to .95 (95%).
verbose: Toggle off warnings.
rvar_col: A single character - the name of an rvar column in the data frame to be processed. See example in p_direction().
effects: Should results for fixed effects, random effects or both be returned? Only applies to mixed models. May be abbreviated.
component: Should results for all parameters, parameters for the conditional model or the zero-inflated part of the model be returned? May be abbreviated. Only applies to brms-models.
parameters: Regular expression pattern that describes the parameters that should be returned. Meta-parameters (like lp__ or prior_) are filtered by default, so only parameters that typically appear in the summary() are returned. Use parameters to select specific parameters for the output.
use_iterations: Logical, if TRUE and x is a get_predicted object, (returned by insight::get_predicted()), the function is applied to the iterations instead of the predictions. This only applies to models that return iterations for predicted values (e.g., brmsfit models).

Value

A data frame with following columns:

Parameter The model parameter(s), if x is a model-object. If x is a vector, this column is missing.
CI The probability of the credible interval.
CI_low, CI_high The lower and upper credible interval limits for the parameters.

Details

The SPI is an alternative method to the HDI (hdi()) to quantify uncertainty of (posterior) distributions. The SPI is said to be more stable than the HDI, because, the "HDI can be noisy (that is, have a high Monte Carlo error)" (Liu et al. 2015). Furthermore, the HDI is sensitive to additional assumptions, in particular assumptions related to the different estimation methods, which can make the HDI less accurate or reliable.

Note

The code to compute the SPI was adapted from the SPIn package, and slightly modified to be more robust for Stan models. Thus, credits go to Ying Liu for the original SPI algorithm and R implementation.

References

Liu, Y., Gelman, A., & Zheng, T. (2015). Simulation-efficient shortest probability intervals. Statistics and Computing, 25(4), 809–819. https://doi.org/10.1007/s11222-015-9563-8

Examples

library(bayestestR)

posterior <- rnorm(1000)
spi(posterior)
#> 95% SPI: [-1.96, 1.96]
spi(posterior, ci = c(0.80, 0.89, 0.95))
#> Shortest Probability Interval
#> 
#> 80% SPI       |       89% SPI |       95% SPI
#> ---------------------------------------------
#> [-1.17, 1.34] | [-1.50, 1.70] | [-1.96, 1.96]

df <- data.frame(replicate(4, rnorm(100)))
spi(df)
#> Shortest Probability Interval
#> 
#> Parameter |       95% SPI
#> -------------------------
#> X1        | [-2.04, 1.89]
#> X2        | [-1.65, 1.96]
#> X3        | [-2.09, 1.74]
#> X4        | [-2.11, 1.97]
spi(df, ci = c(0.80, 0.89, 0.95))
#> Shortest Probability Interval
#> 
#> Parameter |       80% SPI |       89% SPI |       95% SPI
#> ---------------------------------------------------------
#> X1        | [-1.05, 1.49] | [-1.41, 1.76] | [-2.04, 1.89]
#> X2        | [-0.90, 1.09] | [-1.33, 1.41] | [-1.65, 1.96]
#> X3        | [-1.38, 1.44] | [-1.84, 1.49] | [-2.09, 1.74]
#> X4        | [-1.27, 1.40] | [-1.61, 1.56] | [-2.11, 1.97]
# \donttest{
library(rstanarm)
model <- suppressWarnings(
  stan_glm(mpg ~ wt + gear, data = mtcars, chains = 2, iter = 200, refresh = 0)
)
spi(model)
#> Shortest Probability Interval 
#> 
#> Parameter   |        95% SPI
#> ----------------------------
#> (Intercept) | [29.08, 47.70]
#> wt          | [-6.75, -4.12]
#> gear        | [-1.74,  1.70]
# }