Tools to Support the Sheffield Elicitation Framework

Implements various methods for eliciting a probability distribution for a single parameter from an expert or a group of experts. The expert provides a small number of probability judgements, corresponding to points on his or her cumulative distribution function. A range of parametric distributions can then be fitted and displayed, with feedback provided in the form of fitted probabilities and percentiles. A graphical interface for the roulette elicitation method is also provided. For multiple experts, a weighted linear pool can be calculated. Also includes functions for eliciting beliefs about population distributions.


SHELF v1.1.0 (2016-01-29)

  • Change in fitdist to starting values in optimisation: will now check for exact fits if only two probabilities elicited

  • New functions added for eliciting beliefs about uncertain population distributions: cdffeedback, cdfplot, fitprecision, pdfplots

SHELF v1.2.0 (2016-08-16)

  • Roulette elicitation method now implemented using shiny

  • New functions fitDirchlet and feedbackDirichlet for eliciting Dirichlet distributions

  • New functions copulaSample and elicitConcProb for eliciting dependent distributions using multivariate normal copulas

  • New function compareIntervals for comparing fitted intervals for individual distributions from multiple experts

  • Change to expert.names from numbers to letters in fitdist

  • Vignettes added: overview of SHELF, eliciting a Dirichlet distribution, eliciting a bivariate distribution with a bivariate normal copula

SHELF v1.2.1 (2016-09-06)

  • Bug fixed: interactive plots now work for plotting individual distributions for multiple experts

  • Bug fixed: plotting best fitting individual distributions for multiple experts

SHELF v1.2.2 (2016-11-14)

  • Accept non-decreasing probabilities in elicited judgements, rather than only strictly increasing probabilities

  • Can specify own axes labels in the plotfit command with arguments xlab and ylab

  • Update to Multivariate-normal-copula.Rmd vignette, to match update to GGally

SHELF v1.2.3 (2017-02-10)

  • Bug fixed: can fit (and plot) distributions bounded below when lower limit is negative

  • Bug fixed: roulette method shiny interface works with non-integer bin boundaries

SHELF v1.3.0 (2017-10-31)

  • Bug fixed: ensure solid line used for linear pool when plotting. Option in plotfit added to plot all individual densities with same colour, to simplify legend.

  • New function: linearPoolDensity, for extracting density values from the linear pool.

  • Bug fixed: can now accept more than 26 experts.

  • Bug fixed: qlinearpool/plinearpool now works with log t distributions.

  • New function: elicitHeterogen, for eliciting prior for variance of random effects in meta-analysis

Reference manual

It appears you don't have a PDF plugin for this browser. You can click here to download the reference manual.


1.3.0 by Jeremy Oakley, 8 months ago

Report a bug at

Browse source code at

Authors: Jeremy Oakley

Documentation:   PDF Manual  

GPL-2 | GPL-3 license

Imports ggplot2, grid, shiny, stats, graphics, tidyr, MASS, ggExtra

Suggests testthat, knitr, rmarkdown, GGally

See at CRAN