designLink {indirect}

Create list with information for the elicitation session

Description

This builds the structure that will store elicited data. The linear predictor has a normal prior g(\theta) ~ N(m, V), \theta is the elicitation target. Link functions g(.): logit, log, cloglog, identity.

Usage

designLink(
  design,
  link = "identity",
  target = "Target",
  CI.prob = 1/2,
  expertID = "Expert",
  facilitator = "Facilitator",
  rapporteur = "none",
  intro.comments = "This is a record of the elicitation session.",
  fit.method = "KL"
)

Arguments

design

a dataframe with covariate values that will be displayed to the expert(s) during the elicitation session.

link

character logit, log, cloglog, identity

target

character, name of target parameter of elicitation exercise

CI.prob

numeric, a fraction between 0 and 1 that defines probability attributed to central credible interval. For example, 1/2 for a central credible interval of probability 0.5, or 1/3 for a central credible interval of probablity 0.333... The default is probability 1/2.

expertID

character, identifier for expert or group of experts

facilitator

character, facilitator identifier

rapporteur

character, rapporteur identifier. Default "none".

intro.comments

character, text with any prefacing comments. This may include, for example, the definition of the target parameter for the elictation session. Beware of non-ASCII text and special characters, which may affect the ability to save the elicitation record with function saveRecord or create a summary report with function makeSweave if called by the function makeSweave may affect ability to render by means of Sweave or knitr etc.

fit.method

character, method used to fit conditional means prior: KL (default), moment, SS (see vignette and mV for more information on these options)

Details

Assumption: at least two fractiles selected from the median, upper and lower bounds of hte central credible interval of probability CI.prob will be elicited at each design point. The probabilities assigned to the central credible intervals can vary across design points. The argument CI.prob can later be adjusted by design point during the elicitation exercise, see function elicitPt. In the first instance, it is set to a global value specified by CI.prob in function designLink with default value 0.5.

Value

list of design with entries: theta, a n x 4 matrix with columns that give lower, median and upper quantiles followed by CI.prob and n equal to the number of design points (scenarios); link, the link function used; target; expert facilitator; rapporteur; date; intro.comments; fit.method.

Examples

X <- matrix(c(1, 1, 0, 1), nrow = 2) # design
Z <- designLink(design = X, link = "logit", target = "target",
 CI.prob = 1/2, expertID = "Expert", facilitator = "facilitator")