MCMCplot: Caterpillar plots of posterior distributions from MCMC output

In MCMCvis: Tools to Visualize, Manipulate, and Summarize MCMC Output

Caterpillar plots of posterior distributions from MCMC output

Description

Visualize posterior distributions from MCMC output for specific parameters of interest using caterpillar plots. Color of median dot represents the overlap of the posterior distribution with 0 (or other specified value).

Usage

MCMCplot(
  object,
  object2 = NULL,
  params = "all",
  HPD = FALSE,
  ci = c(50, 95),
  excl = NULL,
  ISB = TRUE,
  exact = TRUE,
  ref = 0,
  ref_ovl = FALSE,
  col = "black",
  col2 = "red",
  offset = 0.1,
  rank = FALSE,
  horiz = TRUE,
  xlim,
  ylim,
  xlab,
  ylab,
  main,
  labels,
  guide_lines = FALSE,
  guide_axis = TRUE,
  sz_labels = 1.2,
  sz_med = 1.5,
  sz_thick = 5,
  sz_thin = 2,
  sz_ax = 3,
  sz_ax_txt = 1.3,
  sz_tick_txt = 1.2,
  sz_main_txt = 1.2,
  pos_tick,
  mar = c(5.1, 4.1, 4.1, 2.1)
)

Arguments

object	Object containing MCMC output. See DETAILS below.
object2	Optional second object containing MCMC output. If specified, parameter estimates from each model will be displayed in a paired manner. Parameter names for 'object' and 'object2' must be identical. See DETAILS below.
params	Character string (or vector of character strings) denoting parameters to be plotted. Default 'all' plots posteriors for all parameters. See VALUE below.
HPD	Logical specifying whether to calculate equal-tailed credible intervals (HPD = FALSE) or highest posterior density intervals (HPD = TRUE) for the selected parameters. Default is HPD = FALSE.
ci	Numeric vector of length 2, where each element is (0,100] and represents the width of an equal-tailed (HPD = FALSE) or highest posterior density (HPD = TRUE) credible interval. The first element of this vector corresponds to the thicker (narrower) credible interval displayed on the plot (default is 0.5) and the second element of this vector corresponds to the thinner (wider) credible interval (default is 0.95). The first credible interval width (ci[1]) must be less than or equal to the width of the second credible interval (ci[2]).
excl	Character string (or vector of character strings) denoting parameters to exclude. Used in conjunction with params argument to select parameters of interest.
ISB	Ignore Square Brackets (ISB). Logical specifying whether square brackets should be ignored in the params and excl arguments. If TRUE, square brackets are ignored. If FALSE, square brackets are not ignored. This allows partial names to be used when specifying parameters of interest. Use exact argument to specify whether input from params and excl arguments should be matched exactly.
exact	Logical specifying whether input from params and excl arguments should be matched exactly (after ignoring square brackets if ISB = FALSE). If TRUE, input from params and excl are matched exactly (after taking ISB argument into account). If FALSE, input from params and excl are matched using regular expression format (after taking ISB argument into account).
ref	Value indicating where vertical reference line should be created and what value to use a reference for caterpillar median coloration. Default is ref = 0. Argument NULL will plot no reference line.
ref_ovl	Logical specifying whether the style/color of plotted median dots and CI should be changed based on whether the specified credible intervals (50 % and 95 % by default) overlap the reference line. See DETAILS for more information.
col	Character string (or vector of character strings) specifying which color to render estimates on plot. When ref_ovl = TRUE, this argument has no effect and colors plotted will be based on the credible intervals and reference line. Number of specified colors must equal the number of specified parameters or one.
col2	Character string (or vector of character strings) specifying which color to render estimates on plot for object2 (if specified). Number of specified colors must equal the number of specified parameters or one. Red by default.
offset	Value indicating how much to offset plotted posteriors when object2 is specified (i.e., control the amount of space between the two plotted posteriors for each parameter). The distance from one set of parameters to another corresponds to a value of 1.
rank	Logical specifying whether output should be ranked. If TRUE posteriors will be ranked in decreasing order (based on specified measure of centrality) from top down.
horiz	Logical specifying orientation of plot. If TRUE posteriors will be plotted running horizontally (parallel to the x-axis). If FALSE posteriors will be plotted running vertically (perpendicular to the x-axis).
xlim	Numerical vector of length 2, indicating range of x-axis. Only applicable if horiz = TRUE.
ylim	Numerical vector of length 2, indicating range of y-axis. Only applicable if horiz = FALSE.
xlab	Character string labeling x-axis. Only applicable if horiz = TRUE. Default label is 'Parameter Estimate'. Option NULL will return plot with no label on x-axis.
ylab	Character string labeling y-axis. Only applicable if horiz = FALSE. Default label is 'Parameter Estimate'. Option NULL will return plot with no label on y-axis.
main	Character string indicating title of plot.
labels	Character string (or vector of character strings if plotting > 1 parameter) labeling parameter estimates along y-axis (if horiz = FALSE) or x-axis (if horiz = TRUE). Default option will use parameter names from object. Option NULL will return plot with no labels on axis.
guide_lines	Logical specifying whether to plot reference lines for each parameter in order to better visualize which parameter names correspond to each posterior.
guide_axis	Logical specifying whether a second axis should be plotted (x-axis if HORIZ = TRUE, y-axis if HORIZ = FALSE) to help interpret values on plot.
sz_labels	Number specifying size of text for parameter labels on axis.
sz_med	Number specifying size of points represents posterior medians.
sz_thick	Number specifying thickness of 50 percent CI line (thicker line).
sz_thin	Number specifying thickness of 95 percent CI line (thinner line).
sz_ax	Number specifying thickness of axis and ticks.
sz_ax_txt	Number specifying size of text for axis label.
sz_tick_txt	Number specifying size of text for tick labels on axis.
sz_main_txt	Number specifying size of text for main title.
pos_tick	Numeric vector specifying where ticks on axis should be placed.
mar	Numerical vector of length 4 specifying plot margins - (BOTTOM, LEFT, TOP, RIGHT). Changes to the margin should be made within the function rather than using the par call. Default is c(5.1, 4.1, 4.1, 2.1) - the R plot default.

Details

Points represent posterior medians. Parameters where the smaller specified credible intervals (first value in vector provided to ci argument; 50% by default) overlap 0 (or other value specified by the ref argument) are indicated by 'open' circles and a lighter color than that specified (e.g., gray will be displayed with the default black is specified for col). Parameters where the specific smaller credible intervals DO NOT overlap 0 (or other specified value) AND the larger specified credible intervals (second value in vector provided to ci argument; 95% by default) DO overlap 0 (or other specified value) are indicated by 'closed' circles and a lighter color that that specified. Parameters where the larger specified credible intervals DO NOT overlap 0 (or other specified value) are indicated by 'closed' circles and the color specified (black by default). Thick lines represent the smaller specified credible intervals percent credible intervals (50% by default) while thin lines represent the larger specified credible intervals (95 % by default). ref_ovl = TRUE can be used to enable this feature. When two model objects are supplied to the function (with object and object2) and no argument is supplied to col or col2, light red lines can be interpreted as analogous to gray lines. When ref_ovl = TRUE and a color (or colors) other than the default are specified (with the col and col2 arguments) lighter versions of the color specified are used in black of the light gray and/or light red lines.

When object2 is specified, paired caterpillar plots of each parameter are produced. For this reason, parameter names of object and object2 specified with the params argument must be identical (to be used for comparing posterior estimates of similar models). col and col2 arguments can be specified to change the color of output from object and object2, respectively. By default, output from object is plotted in black and object2 is plotted in red. The ref_ovl argument can also be specified.

object argument can be a stanfit object (rstan package), a CmdStanMCMC object (cmdstanr package), a stanreg object (rstanarm package), a brmsfit object (brms package), an mcmc.list object (coda and rjags packages), mcmc object (coda and nimble packages), list object (nimble package), an R2jags model object (R2jags package), a jagsUI model object (jagsUI package), or a matrix containing MCMC chains (each column representing MCMC output for a single parameter, rows representing iterations in the chain). The function automatically detects the object type and proceeds accordingly.

Notes

When specifying rank = TRUE and specifying labels for labels, labels will be applied to parameters before they are ranked.

Thanks to Cinner et al. 2016, whose Fig. 1 inspired this plot.

References

Cinner, J. E., C. Huchery, M. A. MacNeil, N. A. J. Graham, T. R. McClanahan, J. Maina, E. Maire, J. N. Kittinger, C. C. Hicks, C. Mora, E. H. Allison, S. D'Agata, A. Hoey, D. A. Feary, L. Crowder, I. D. Williams, M. Kulbicki, L. Vigliola, L. Wantiez, G. Edgar, R. D. Stuart-Smith, S. A. Sandin, A. L. Green, M. J. Hardt, M. Beger, A. Friedlander, S. J. Campbell, K. E. Holmes, S. K. Wilson, E. Brokovich, A. J. Brooks, J. J. Cruz-Motta, D. J. Booth, P. Chabanet, C. Gough, M. Tupper, S. C. A. Ferse, U. R. Sumaila, and D. Mouillot. 2016. Bright spots among the world's coral reefs. Nature 535:416-419.

Examples

#Load data
data(MCMC_data)

#Plot MCMC output
MCMCplot(MCMC_data, labels = NULL)

#Just 'beta' parameters
MCMCplot(MCMC_data, params = 'beta')

#Just 'beta' parameters using highest posterior density intervals
MCMCplot(MCMC_data, params = 'beta', HPD = TRUE)
#Just 'beta[1]', 'beta[4]', and 'alpha[3]'
MCMCplot(MCMC_data, params = c('beta[1]', 'beta[4]', 'alpha[3]'), ISB = FALSE, exact = TRUE)

#Just 'beta[1]', 'beta[4]', and 'alpha[3]' and change the credible interval widths
MCMCplot(MCMC_data, ci = c(50, 89), params = c('beta[1]', 'beta[4]', 'alpha[3]'), 
  ISB = FALSE, exact = TRUE)

#Rank parameters by posterior mean
MCMCplot(MCMC_data, params = 'beta', rank = TRUE)

#Create vertical plot
MCMCplot(MCMC_data, params = 'beta', horiz = FALSE)

MCMCvis documentation built on Oct. 18, 2023, 1:10 a.m.