Skip to contents

Setting priors

Source: R/set_priors.R
prior_setting.Rd

Functionality for altering priors:

make_priors Generates priors for a model.

set_priors Adds priors to a model.

Extracts priors as a named vector

Usage

make_priors(
  model,
  alphas = NA,
  distribution = NA,
  alter_at = NA,
  node = NA,
  nodal_type = NA,
  label = NA,
  param_set = NA,
  given = NA,
  statement = NA,
  join_by = "|",
  param_names = NA
)

set_priors(
  model,
  alphas = NA,
  distribution = NA,
  alter_at = NA,
  node = NA,
  nodal_type = NA,
  label = NA,
  param_set = NA,
  given = NA,
  statement = NA,
  join_by = "|",
  param_names = NA
)

get_priors(model, nodes = NULL)

Arguments

model

A model object generated by make_model().

alphas

Real positive numbers giving hyperparameters of the Dirichlet distribution

distribution

string indicating a common prior distribution (uniform, jeffreys or certainty)

alter_at

string specifying filtering operations to be applied to parameters_df, yielding a logical vector indicating parameters for which values should be altered. (see examples)

node

string indicating nodes which are to be altered

nodal_type

string. Label for nodal type indicating nodal types for which values are to be altered

label

string. Label for nodal type indicating nodal types for which values are to be altered. Equivalent to nodal_type.

param_set

string indicating the name of the set of parameters to be altered

given

string indicates the node on which the parameter to be altered depends

statement

causal query that determines nodal types for which values are to be altered

join_by

string specifying the logical operator joining expanded types when statement contains wildcards. Can take values '&' (logical AND) or '|' (logical OR).

param_names

vector of strings. The name of specific parameter in the form of, for example, 'X.1', 'Y.01'

nodes

a vector of nodes

Value

A vector indicating the parameters of the prior distribution of the nodal types ("hyperparameters").

An object of class causal_model. It essentially returns a list containing the elements comprising a model (e.g. 'statement', 'nodal_types' and 'DAG') with the `priors` attached to it.

A vector indicating the hyperparameters of the prior distribution of the nodal types.

Details

Seven arguments govern which parameters should be altered. The default is 'all' but this can be reduced by specifying

* alter_at String specifying filtering operations to be applied to parameters_df, yielding a logical vector indicating parameters for which values should be altered. "node == 'X' & nodal_type

* node, which restricts for example to parameters associated with node 'X'

* label or nodal_type The label of a particular nodal type, written either in the form Y0000 or Y.Y0000

* param_set The param_set of a parameter.

* given Given parameter set of a parameter.

* statement, which restricts for example to nodal types that satisfy the statement 'Y[X=1] > Y[X=0]'

* param_set, given, which are useful when setting confound statements that produce several sets of parameters

Two arguments govern what values to apply:

* alphas is one or more non-negative numbers and

* distribution indicates one of a common class: uniform, Jeffreys, or 'certain'

Forbidden statements include:

  • Setting distribution and values at the same time.

  • Setting a distribution other than uniform, Jeffreys, or certainty.

  • Setting negative values.

  • specifying alter_at with any of node, nodal_type, param_set, given, statement, or param_names

  • specifying param_names with any of node, nodal_type, param_set, given, statement, or alter_at

  • specifying statement with any of node or nodal_type

Examples


# make_priors examples:

# Pass all nodal types
model <- make_model("Y <- X")
make_priors(model, alphas = .4)
#>  X.0  X.1 Y.00 Y.10 Y.01 Y.11 
#>  0.4  0.4  0.4  0.4  0.4  0.4 
make_priors(model, distribution = "jeffreys")
#> Altering all parameters.
#>  X.0  X.1 Y.00 Y.10 Y.01 Y.11 
#>  0.5  0.5  0.5  0.5  0.5  0.5 

model <- CausalQueries::make_model("X -> M -> Y; X <-> Y")

#altering values using \code{alter_at}
make_priors(model = model, alphas = c(0.5,0.25),
alter_at = "node == 'Y' & nodal_type %in% c('00','01') & given == 'X.0'")
#>      X.0      X.1     M.00     M.10     M.01     M.11 Y.00_X.0 Y.10_X.0 
#>     1.00     1.00     1.00     1.00     1.00     1.00     0.50     1.00 
#> Y.01_X.0 Y.11_X.0 Y.00_X.1 Y.10_X.1 Y.01_X.1 Y.11_X.1 
#>     0.25     1.00     1.00     1.00     1.00     1.00 

#altering values using \code{param_names}
make_priors(model = model, alphas = c(0.5,0.25),
param_names = c("Y.10_X.0","Y.10_X.1"))
#>      X.0      X.1     M.00     M.10     M.01     M.11 Y.00_X.0 Y.10_X.0 
#>     1.00     1.00     1.00     1.00     1.00     1.00     1.00     0.50 
#> Y.01_X.0 Y.11_X.0 Y.00_X.1 Y.10_X.1 Y.01_X.1 Y.11_X.1 
#>     1.00     1.00     1.00     0.25     1.00     1.00 

#altering values using \code{statement}
make_priors(model = model, alphas = c(0.5,0.25),
statement = "Y[M=1] > Y[M=0]")
#> Warning: Possible ambiguity: use additional arguments or check behavior in parameters_df.
#>      X.0      X.1     M.00     M.10     M.01     M.11 Y.00_X.0 Y.10_X.0 
#>     1.00     1.00     1.00     1.00     1.00     1.00     1.00     1.00 
#> Y.01_X.0 Y.11_X.0 Y.00_X.1 Y.10_X.1 Y.01_X.1 Y.11_X.1 
#>     0.50     1.00     1.00     1.00     0.25     1.00 

#altering values using a combination of other arguments
make_priors(model = model, alphas = c(0.5,0.25),
node = "Y", nodal_type = c("00","01"), given = "X.0")
#>      X.0      X.1     M.00     M.10     M.01     M.11 Y.00_X.0 Y.10_X.0 
#>     1.00     1.00     1.00     1.00     1.00     1.00     0.50     1.00 
#> Y.01_X.0 Y.11_X.0 Y.00_X.1 Y.10_X.1 Y.01_X.1 Y.11_X.1 
#>     0.25     1.00     1.00     1.00     1.00     1.00 

# set_priors examples:

# Pass all nodal types
model <- make_model("Y <- X")
set_priors(model, alphas = .4)
#> 
#> Causal statement: 
#> X -> Y
#> 
#> Number of types by node:
#> X Y 
#> 2 4 
#> 
#> Number of causal types: 8
set_priors(model, distribution = "jeffreys")
#> Altering all parameters.
#> 
#> Causal statement: 
#> X -> Y
#> 
#> Number of types by node:
#> X Y 
#> 2 4 
#> 
#> Number of causal types: 8

model <- CausalQueries::make_model("X -> M -> Y; X <-> Y")

#altering values using \code{alter_at}
set_priors(model = model, alphas = c(0.5,0.25),
alter_at = "node == 'Y' & nodal_type %in% c('00','01') & given == 'X.0'")
#> 
#> Causal statement: 
#> M -> Y; X -> M; X <-> Y
#> 
#> Number of types by node:
#> X M Y 
#> 2 4 4 
#> 
#> Number of causal types: 32

#altering values using \code{param_names}
set_priors(model = model, alphas = c(0.5,0.25),
param_names = c("Y.10_X.0","Y.10_X.1"))
#> 
#> Causal statement: 
#> M -> Y; X -> M; X <-> Y
#> 
#> Number of types by node:
#> X M Y 
#> 2 4 4 
#> 
#> Number of causal types: 32

#altering values using \code{statement}
set_priors(model = model, alphas = c(0.5,0.25),
statement = "Y[M=1] > Y[M=0]")
#> Warning: Possible ambiguity: use additional arguments or check behavior in parameters_df.
#> 
#> Causal statement: 
#> M -> Y; X -> M; X <-> Y
#> 
#> Number of types by node:
#> X M Y 
#> 2 4 4 
#> 
#> Number of causal types: 32

#altering values using a combination of other arguments
set_priors(model = model, alphas = c(0.5,0.25), node = "Y",
nodal_type = c("00","01"), given = "X.0")
#> 
#> Causal statement: 
#> M -> Y; X -> M; X <-> Y
#> 
#> Number of types by node:
#> X M Y 
#> 2 4 4 
#> 
#> Number of causal types: 32