Nothing
R/sanitization-fitted-assignment.R
In bnlearn: Bayesian Network Structure Learning, Parameter Learning and Inference
Defines functions
check.discrete.parents.configuration
check.cgnode.rvalue.vs.parents
check.rvalue.vs.cgnode
check.cgnode.rvalue
check.gnode.rvalue.vs.parents
check.rvalue.vs.gnode
check.gnode.rvalue
check.dnode.rvalue.vs.parents
check.rvalue.vs.dnode
check.dnode.rvalue
# check the consistency of an assignment to a bn.fit.dnode or a bn.fit.onode.
check.dnode.rvalue = function(x, node) {
# the CPT must be a table.
if (!is.ndmatrix(x))
stop("the conditional probability distribution of node ", node,
" must be a table, a matrix or a multidimensional array.")
# all elements must be probabilities.
if (!is.probability.vector(x))
stop("the probabilities provided for node ",
node, " do not form a valid conditional probability distribution.")
# convert the CPT into a table object.
x = as.table(x)
# compute the dimensions of the table.
dims = dim(x)
ndims = length(dims)
# flatten 1xc tables into 1-dimensional ones.
if ((ndims == 2) && (dims[1] == 1)) {
x = as.table(structure(c(x), names = colnames(x)))
names(dimnames(x)) = ""
}#THEN
# update dims and ndims.
ndims = length(dim(x))
# one dimension name in a multi-dimensional table must be the node label.
dnn = dimnames(x)
if (ndims >= 2)
if (all(is.character(names(dnn))) && all(names(dnn) != node))
stop("node ", node, " does not have a corresponding dimension name.")
# do not allow missing values as dimension names.
nas = sapply(dnn, anyNA)
if (any(nas))
stop("node ", node, " contains missing values in the dimension names.")
# do not allow duplicated dimension names.
dupes = sapply(dnn, anyDuplicated)
if (any(dupes > 0))
stop("node ", node, " contains duplicate dimension names.")
# normalization.
if (ndims == 1) {
if (abs(sum(x) - 1) > 0.01)
stop("the probability distribution of node ", node, " does not sum to one.")
x = x / sum(x)
}#THEN
else {
# if the new table has dimensions names, use them to decide over which
# dimension the probabilities should sum up to 1; otherwise assume that
# probabilities sum up to 1 over rows.
if (node %in% names(dnn))
normalize.over = which(names(dnn) != node)
else
normalize.over = seq(from = 2, to = ndims)
fails.check.sum = sapply(margin.table(x, normalize.over),
function(x) abs(sum(x) - 1) > 0.01)
if (any(fails.check.sum))
stop("some conditional probability distributions of node ", node,
" do not sum to one.")
x = prop.table(x, margin = normalize.over)
}#ELSE
return(x)
}#CHECK.DNODE.RVALUE
# check an assignment against the bn.fit.dnode or bn.fit.onode it will be
# assigned to.
check.rvalue.vs.dnode = function(new, old) {
ndims = length(dim(new))
if (!is(old, c("bn.fit.dnode", "bn.fit.onode")))
stop("both nodes should be discrete nodes.")
# now that we are sure that the dimensions are the right ones, reorder
# them to follow the ordering of the parents in the network.
ndnn = dimnames(new)
odnn = dimnames(old$prob)
if (!is.null(ndnn)) {
# if the new CPT has dimension names, they must match (and be in the same
# order too).
if (ndims > 1) {
if (!setequal(names(odnn), names(ndnn)))
stop("wrong dimension names for node ", old$node, ".")
new = aperm(new, match(names(odnn), names(ndnn)))
ndnn = dimnames(new)
}#THEN
# check labels in the same way, if present.
if (any(!sapply(ndnn, is.null))) {
dnn.check = sapply(seq_along(ndnn),
function(i) setequal(ndnn[[i]], odnn[[i]]))
if (!all(dnn.check))
stop("wrong levels for node ", old$node, ".")
new = ndsubset(new, odnn)
}#THEN
else {
if (!identical(dim(new), dim(old$prob)))
stop("wrong dimensions for node ", old$node, ".")
}#ELSE
}#THEN
else {
# same dimensions (there is nothing else to check).
if (!identical(dim(new), dim(old$prob)))
stop("wrong dimensions for node ", old$node, ".")
}#ELSE
# make sure the returned object is a table.
return(structure(new, class = "table"))
}#CHECK.RVALUE.VS.DNODE
# check an assignment to a bn.fit.dnode or a bn.fit.onode against the local
# distributions of its parents.
check.dnode.rvalue.vs.parents = function(node, new, parents) {
if (length(parents) == 0)
return(new)
# all parents of discrete nodes must be discrete nodes themselves.
illegal.parents =
sapply(parents, class) %!in% c("bn.fit.dnode", "bn.fit.onode")
if (any(illegal.parents))
stop("node ", node, " is discrete but has continuous parents (",
names(parents)[illegal.parents], ").")
# extract the CPTs.
cpt.levels = dimnames(new)
parents.levels = smartSapply(NULL, names(parents),
function(p) dimnames(parents[[p]]$prob)[[1]])
# check the number of dimensions.
if (length(cpt.levels) != length(parents) + 1)
stop("wrong number of parents for node ", node, ".")
# check the dimension names, and add them if they are missing
if (is.null(names(cpt.levels)))
names(dimnames(new)) = names(cpt.levels) = c(node, names(parents.levels))
else {
# check whether the CPT covers all the relevant variables.
if (!setequal(names(cpt.levels)[-1], names(parents.levels)))
stop("wrong dimensions for node ", node, ".")
# now that we are sure that the dimensions are the right ones, reorder
# them to follow the ordering of the parents in the network.
d = names(dimnames(new))
new = aperm(new, c(match(node, d), match(names(parents.levels), d)))
cpt.levels = dimnames(new)
}#ELSE
# check the levels in the CPT against those of the parents, and reoder them
# if needed.
for (p in names(parents))
if (!setequal(parents.levels[[p]], cpt.levels[[p]]))
stop("the levels of the parent ", p, " of node ", node, " do not match.")
if (!isTRUE(all.equal(parents.levels, cpt.levels[-1])))
new = ndsubset(new, parents.levels)
return(new)
}#CHECK.DNODE.RVALUE.VS.PARENTS
# check the consistency of an assignment to a bn.fit.gnode.
check.gnode.rvalue = function(x, node) {
components = c("coef", "fitted", "resid", "sd")
labels = c(fitted = "fitted values", resid = "residuals")
# custom list of components.
if (!is.list(x) || any(names(x) %!in% components))
stop("the conditional probability distribution for node ", node,
" must be a list with at least one of the following elements:",
paste0(" '", components, "'"), ".")
if (("coef" %!in% names(x)) || !any(c("sd", "resid") %in% names(x)))
stop("at least the regression coefficients and either the residuals or ",
"the residual standard deviation are required for node ", node, ".")
if (!is.null(x$coef))
if ((length(x$coef) == 0) || !is.real.vector(x$coef))
stop("coef must be a vector of numeric values, the ",
"regression coefficients for node ", node, " given its parents.")
# check that the coefficients are stored as numeric values, not integers.
x$coef[] = as.numeric(x$coef)
for (comp in c("fitted", "resid"))
if (!is.null(x[[comp]]))
if ((length(x[[comp]]) == 0) || !is.real.vector(x[[comp]]))
stop(comp, " must be a vector of numeric values, the ",
labels[comp], " for node ", node, " given its parents.")
# check the standard deviation of the residuals.
if (!is.null(x$sd)) {
if (!is.non.negative(x$sd))
stop("sd must be a non-negative scalar, the standard deviation of the ",
"residuals of node ", node, ".")
if (!is.null(x$resid)) {
adj.sd = cgsd(x$resid, p = length(x$coef))
if (!isTRUE(all.equal(x$sd, adj.sd, check.attributes = FALSE, tol = 0.0005)))
stop("the reported standard deviation of the residuals of node ", node,
" does not match the observed one.")
}#THEN
}#THEN
else if (!is.null(x$resid)) {
# compute it from the residuals if not provided, it must be defined.
x$sd = cgsd(x$resid, configs = x$configs, p = length(x$coef))
}#THEN
# make sure that the standard error is not an integer variable.
storage.mode(x$sd) = "double"
# one residual for each fitted value.
if (!is.null(x$resid) && !is.null(x$fitted))
if (length(x$resid) != length(x$fitted))
stop("the residuals and the fitted values of node ", node,
" have different lengths.")
return(x)
}#CHECK.GNODE.RVALUE
# check an assignment against the bn.fit.gnode it will be assigned to.
check.rvalue.vs.gnode = function(new, old) {
if (!is(old, "bn.fit.gnode"))
stop("both nodes should be continuous (Gaussian) nodes.")
# same number of coefficients.
if (length(new$coef) != length(old$coefficients))
stop("wrong number of coefficients for node ", old$node, ".")
# same regression coefficients.
if (!is.null(names(new$coef))) {
# if the new coefficients have labels, they must match.
check.nodes(names(new$coef), graph = names(old$coefficients),
min.nodes = length(names(old$coefficients)))
# reorder the coefficients to match.
new$coef = new$coef[names(old$coefficients)]
}#THEN
else {
# copy the names from the spec.
names(new$coef) = names(old$coefficients)
}#ELSE
# same number of standard errors (fixed to 1).
if (length(new$sd) != 1)
stop("wrong number of standard errors for node ", old$node, ".")
# same number of residuals.
if (!is.null(new$resid))
if (length(new$resid) != length(old$residuals))
stop("wrong number of residuals for node ", old$node, ".")
# same number of fitted values.
if (!is.null(new$fitted))
if (length(new$fitted) != length(old$fitted.values))
stop("wrong number of fitted values for node ", old$node, ".")
return(new)
}#CHECK.RVALUE.VS.GNODE
# check an assignment to a bn.fit.gnode using only the names of its parents.
check.gnode.rvalue.vs.parents = function(node, new, parents) {
# all parents of Gaussian nodes must be continuous nodes.
illegal.parents =
sapply(parents, class) %!in% c("bn.fit.gnode", "bn.fit.cgnode")
if (any(illegal.parents))
stop("node ", node, " is Gaussian but has discrete parents (",
names(parents)[illegal.parents], ").")
# add the intercept, which is obviously not among the parents of the node.
parent.names = c("(Intercept)", names(parents))
# same number of coefficients.
if (length(new$coef) != length(parent.names))
stop("wrong number of coefficients for node ", node, ".")
# same number of standard errors (fixed to 1).
if (length(new$sd) != 1)
stop("wrong number of standard errors for node ", node, ".")
# if the new coefficients have labels, they must match.
if (!is.null(names(new$coef))) {
if (!setequal(names(new$coef), parent.names))
stop("wrong regression coefficients for node ", node, " (",
paste(setdiff(parent.names, names(new$coef)), collapse = " "), ").")
new$coef = new$coef[parent.names]
}#THEN
else {
names(new$coef) = parent.names
}#ELSE
return(new)
}#CHECK.GNODE.RVALUE.VS.PARENTS
# check the consistency of an assignment to a bn.fit.cgnode.
check.cgnode.rvalue = function(x, node) {
components = c("coef", "fitted", "resid", "sd", "configs")
labels = c(fitted = "fitted values", resid = "residuals")
# custom list of components.
if (!is.list(x) || any(names(x) %!in% components))
stop("the conditional probability distribution for node ", node,
" must be a list with at least one of the following elements:",
paste0(" '", components, "'"), ".")
if (("coef" %!in% names(x)) || !any(c("sd", "resid") %in% names(x)))
stop("at least the regression coefficients and either the residuals or ",
"the residual standard deviation are required for node ", node, ".")
if (!is.null(x$coef))
if ((length(x$coef) == 0) || !is.real.vector(x$coef))
stop("coef must be a matrix of numeric values, the ",
"regression coefficients for node ", node, " given its parents.")
# check that the coefficients are stored as numeric values, not integers.
x$coef[] = as.numeric(x$coef)
for (comp in c("fitted", "resid"))
if (!is.null(x[[comp]]))
if ((length(x[[comp]]) == 0) || !is.real.vector(x[[comp]]))
stop(comp, " must be a vector of numeric values, the ",
labels[comp], " for node ", node, " given its parents.")
if (!is.null(x$configs)) {
if ((length(x$configs) == 0) || !is.factor(x$configs))
stop("the discrete parents' configurations for node ", node,
" must be a factor.")
if (nlevels(x$config) != ncol(x$coef))
stop("wrong number of discrete parents' configurations for node ", node, ".")
}#THEN
# check the standard deviations of the residuals.
if (!is.null(x$sd)) {
if (!is.nonnegative.vector(x$sd))
stop("sd must be non-negative, the standard deviations of the ",
"residuals of node ", node, ".")
if (length(x$sd) != ncol(x$coef))
stop("the dimensions of sd and coef do not match in node ", node, ".")
if (!is.null(x$resid) && !is.null(x$configs)) {
adj.sd = cgsd(x$resid, configs = x$configs, p = nrow(x$coef))
if (!isTRUE(all.equal(x$sd, adj.sd, check.attributes = FALSE, tol = 0.0005)))
stop("the reported standard deviations of the residuals of node ", node,
" does not match the observed one.")
}#THEN
}#THEN
else if (!is.null(x$resid)) {
# compute the standard error from the residuals if possible.
if (is.null(x$config))
stop("sd is missing, and parents' configurations are required to compute it.")
x$sd = cgsd(x$resid, configs = x$configs,
p = ifelse(is.matrix(x$coef), nrow(x$coef), length(x$coef)))
}#THEN
# make sure that the standard error is not an integer variable.
storage.mode(x$sd) = "double"
# one residual for each fitted value.
if (!is.null(x$resid) && !is.null(x$fitted))
if (length(x$resid) != length(x$fitted))
stop("the residuals and the fitted values of node ", node,
" have different lengths.")
# if any, one parent configuration for each fitted value and residual.
if (!is.null(x$config)) {
if (!is.null(x$resid) && (length(x$configs) != length(x$resid)))
stop("parents' configurations and residuals of node ", node,
" have different lengths.")
if (!is.null(x$fitted) && (length(x$configs) != length(x$fitted)))
stop("parents' configurations and fitted values of node ", node,
" have different lengths.")
}#THEN
return(x)
}#CHECK.CGNODE.RVALUE
# check an assignment against the bn.fit.cgnode it will be assigned to.
check.rvalue.vs.cgnode = function(new, old) {
if (!is(old, "bn.fit.cgnode"))
stop("both nodes should be continuous (conditional Gaussian) nodes.")
# right dimensions for the coefficients.
if (!is(new$coef, "matrix") ||
!identical(dim(new$coef), dim(old$coefficients)))
stop("the regression coefficients for node ", old$node, " must be ",
"in a matrix with one column for each discrete parent and one ",
"coefficient for each continuous parent.")
# right row names for the coefficients.
if (!is.null(rownames(new$coef))) {
# if the new coefficients have labels, they must match.
check.nodes(rownames(new$coef), graph = rownames(old$coefficients),
min.nodes = length(rownames(old$coefficients)))
# reorder the coefficients to match.
new$coef = new$coef[rownames(old$coefficients), , drop = FALSE]
}#THEN
else {
# copy the names from the spec.
rownames(new$coef) = rownames(old$coefficients)
}#ELSE
# right column names for the coefficients.
if (!is.null(colnames(new$coef))) {
# if the new coefficients have labels, they must match.
check.nodes(colnames(new$coef), graph = colnames(old$coefficients),
min.nodes = length(colnames(old$coefficients)))
# reorder the coefficients to match.
new$coef = new$coef[, colnames(old$coefficients), drop = FALSE]
}#THEN
else {
# copy the names from the spec.
colnames(new$coef) = colnames(old$coefficients)
}#ELSE
# same number of residuals.
if (!is.null(new$resid))
if (length(new$resid) != length(old$residuals))
stop("wrong number of residuals for node ", old$node, ".")
# same number of fitted values.
if (!is.null(new$fitted))
if (length(new$fitted) != length(old$fitted.values))
stop("wrong number of fitted values for node ", old$node, ".")
if (!is.null(new$dlevels)) {
old.dparents = names(old$dlevels)
new.dparents = names(new$dlevels)
# different sets of discrete parents.
if (!setequal(old.dparents, new.dparents))
stop("different discrete parents for node ", old$node, ".")
# identical sets of discrete parents, but in a different order; match
# configurations.
old.config = old$dlevels
new.config = new$dlevels
for (i in seq_along(old.config))
old.config[[i]] = paste(old.dparents[i], old.config[[i]], sep = ":")
for (i in seq_along(new.config))
new.config[[i]] = paste(new.dparents[i], new.config[[i]], sep = ":")
old.config = interaction(expand.grid(old.config))
new.config = interaction(expand.grid(new.config)[old.dparents])
map = match(old.config, new.config)
# reorder coefficients, standard errors and configurations.
new$sd = new$sd[map]
new$coef = new$coef[, map, drop = FALSE]
new$dlevels = new$dlevels[old.dparents]
colnames(new$coef) = names(new$sd) = names(old$sd)
}#THEN
return(new)
}#CHECK.RVALUE.VS.CGNODE
# check an assignment to a bn.fit.cgnode against the local distributions of its
# parents.
check.cgnode.rvalue.vs.parents = function(node, new, parents) {
# both continuous and discrete parents are legal, but treated differently.
discrete.id = sapply(parents, class) %in% c("bn.fit.dnode", "bn.fit.onode")
discrete.parents = names(parents)[discrete.id]
continuous.id = sapply(parents, class) %in% c("bn.fit.gnode", "bn.fit.cgnode")
continuous.parents = names(parents)[continuous.id]
parent.configurations =
prod(sapply(parents[discrete.parents], function(p) dim(p$prob)[1]))
correct.configs = as.character(seq(from = 0, to = parent.configurations - 1))
new$dparents = which(discrete.id)
new$gparents = which(continuous.id)
new$dlevels = lapply(parents[discrete.parents], function(p) dimnames(p$prob)[[1]])
# the node should have at least one discrete parent, or it would not be CLG.
if (length(discrete.parents) == 0)
stop("node ", node, " is a conditioanl Gaussian node but has no discrete parent.")
# check the dimensions of the regression coefficients matrix.
if (!isTRUE(all.equal(dim(new$coef), c(length(continuous.parents) + 1L,
parent.configurations))))
stop("wrong dimensions for the regression coefficients matrix of node ",
node, ".")
# check and fix the coefficients names.
if (!is.null(rownames(new$coef))) {
regression.coefficients = c("(Intercept)", continuous.parents)
if (!setequal(rownames(new$coef), regression.coefficients))
stop("wrong regression coefficients for node ", node, " (",
paste(setdiff(continuous.parents, rownames(new$coef)), collapse = " "), ").")
new$coef = new$coef[regression.coefficients, , drop = FALSE]
}#THEN
else {
rownames(new$coef) = c("(Intercept)", continuous.parents)
}#ELSE
# check and fix the column names of the coefficients matrix.
if (is.null(colnames(new$coef)))
colnames(new$coef) = correct.configs
else if (setequal(colnames(new$coef), correct.configs))
new$coef = new$coef[, correct.configs, drop = FALSE]
else {
# wrong names: warn and rename.
warning("remapping levels of the discrete parents configurations ",
"for node ", node, ".")
colnames(new$coef) = correct.configs
}#ELSE
# check the number of standard errors, and check their names.
if (length(new$sd) != parent.configurations)
stop("wrong number of standard errors for node ", node, ".")
if (is.null(names(new$sd)))
names(new$sd) = correct.configs
else if (setequal(names(new$sd), correct.configs))
new$sd = new$sd[correct.configs]
else {
# wrong names: warn and rename.
warning("remapping levels of the discrete parents configurations ",
"for node ", node, ".")
names(new$sd) = correct.configs
}#ELSE
# check the discrete parents configurations.
if (!is.null(new$configs)) {
actual.configs = levels(new$configs)
if (length(actual.configs) != parent.configurations)
stop("wrong number of discrete parents' configurations for node", node, ".")
if (any(actual.configs != correct.configs)) {
if (setequal(actual.configs, correct.configs))
new$configs = factor(new$configs, correct.configs)
else {
# wrong levels: warn and rename.
warning("remapping levels of the discrete parents configurations ",
"for node ", node, ".")
levels(new$configs) = correct.configs
}#ELSE
}#THEN
}#THEN
return(new)
}#CHECK.CGNODE.RVALUE.VS.PARENTS
# check discrete parents configurations in list form for all kinds of nodes.
check.discrete.parents.configuration = function(config, node,
ideal.only = FALSE) {
# check whether the configuration is there.
if (missing(config) || !is(config, "list"))
stop("the parents configuration must be a list with elements named after the parents of ", node$node, ".")
if (is(node, c("bn.fit.dnode", "bn.fit.onode")))
allowed.parents = node$parents
else if (is(node, c("bn.fit.cgnode")))
allowed.parents = node$parents[node$dparents]
# check the node labels from the list names.
check.nodes(names(config), graph = allowed.parents,
min.nodes = length(allowed.parents), max.nodes = length(allowed.parents))
# check that only one value is provided for each parent, to get a single
# configuration.
for (c in names(config))
if (length(config[[c]]) > 1)
stop("only one value allowed for parent '", c, "'.")
if (is(node, c("bn.fit.dnode", "bn.fit.onode")))
levels = dimnames(node$prob)
else
levels = node$dlevels
# check whether the provided values are valid values.
for (c in names(config)) {
if (is.factor(config[[c]]))
config[[c]] = as.character(config[[c]])
if (config[[c]] %!in% levels[[c]])
stop("the value of parent '", c, "' must be a valid level.")
}#FOR
return(config)
}#CHECK.DISCRETE.PARENTS.CONFIGURATION
Try the bnlearn package in your browser
Any scripts or data that you put into this service are public.
bnlearn documentation built on Sept. 11, 2024, 8:27 p.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Defines functions check.discrete.parents.configuration check.cgnode.rvalue.vs.parents check.rvalue.vs.cgnode check.cgnode.rvalue check.gnode.rvalue.vs.parents check.rvalue.vs.gnode check.gnode.rvalue check.dnode.rvalue.vs.parents check.rvalue.vs.dnode check.dnode.rvalue
# check the consistency of an assignment to a bn.fit.dnode or a bn.fit.onode.
check.dnode.rvalue = function(x, node) {
# the CPT must be a table.
if (!is.ndmatrix(x))
stop("the conditional probability distribution of node ", node,
" must be a table, a matrix or a multidimensional array.")
# all elements must be probabilities.
if (!is.probability.vector(x))
stop("the probabilities provided for node ",
node, " do not form a valid conditional probability distribution.")
# convert the CPT into a table object.
x = as.table(x)
# compute the dimensions of the table.
dims = dim(x)
ndims = length(dims)
# flatten 1xc tables into 1-dimensional ones.
if ((ndims == 2) && (dims[1] == 1)) {
x = as.table(structure(c(x), names = colnames(x)))
names(dimnames(x)) = ""
}#THEN
# update dims and ndims.
ndims = length(dim(x))
# one dimension name in a multi-dimensional table must be the node label.
dnn = dimnames(x)
if (ndims >= 2)
if (all(is.character(names(dnn))) && all(names(dnn) != node))
stop("node ", node, " does not have a corresponding dimension name.")
# do not allow missing values as dimension names.
nas = sapply(dnn, anyNA)
if (any(nas))
stop("node ", node, " contains missing values in the dimension names.")
# do not allow duplicated dimension names.
dupes = sapply(dnn, anyDuplicated)
if (any(dupes > 0))
stop("node ", node, " contains duplicate dimension names.")
# normalization.
if (ndims == 1) {
if (abs(sum(x) - 1) > 0.01)
stop("the probability distribution of node ", node, " does not sum to one.")
x = x / sum(x)
}#THEN
else {
# if the new table has dimensions names, use them to decide over which
# dimension the probabilities should sum up to 1; otherwise assume that
# probabilities sum up to 1 over rows.
if (node %in% names(dnn))
normalize.over = which(names(dnn) != node)
else
normalize.over = seq(from = 2, to = ndims)
fails.check.sum = sapply(margin.table(x, normalize.over),
function(x) abs(sum(x) - 1) > 0.01)
if (any(fails.check.sum))
stop("some conditional probability distributions of node ", node,
" do not sum to one.")
x = prop.table(x, margin = normalize.over)
}#ELSE
return(x)
}#CHECK.DNODE.RVALUE
# check an assignment against the bn.fit.dnode or bn.fit.onode it will be
# assigned to.
check.rvalue.vs.dnode = function(new, old) {
ndims = length(dim(new))
if (!is(old, c("bn.fit.dnode", "bn.fit.onode")))
stop("both nodes should be discrete nodes.")
# now that we are sure that the dimensions are the right ones, reorder
# them to follow the ordering of the parents in the network.
ndnn = dimnames(new)
odnn = dimnames(old$prob)
if (!is.null(ndnn)) {
# if the new CPT has dimension names, they must match (and be in the same
# order too).
if (ndims > 1) {
if (!setequal(names(odnn), names(ndnn)))
stop("wrong dimension names for node ", old$node, ".")
new = aperm(new, match(names(odnn), names(ndnn)))
ndnn = dimnames(new)
}#THEN
# check labels in the same way, if present.
if (any(!sapply(ndnn, is.null))) {
dnn.check = sapply(seq_along(ndnn),
function(i) setequal(ndnn[[i]], odnn[[i]]))
if (!all(dnn.check))
stop("wrong levels for node ", old$node, ".")
new = ndsubset(new, odnn)
}#THEN
else {
if (!identical(dim(new), dim(old$prob)))
stop("wrong dimensions for node ", old$node, ".")
}#ELSE
}#THEN
else {
# same dimensions (there is nothing else to check).
if (!identical(dim(new), dim(old$prob)))
stop("wrong dimensions for node ", old$node, ".")
}#ELSE
# make sure the returned object is a table.
return(structure(new, class = "table"))
}#CHECK.RVALUE.VS.DNODE
# check an assignment to a bn.fit.dnode or a bn.fit.onode against the local
# distributions of its parents.
check.dnode.rvalue.vs.parents = function(node, new, parents) {
if (length(parents) == 0)
return(new)
# all parents of discrete nodes must be discrete nodes themselves.
illegal.parents =
sapply(parents, class) %!in% c("bn.fit.dnode", "bn.fit.onode")
if (any(illegal.parents))
stop("node ", node, " is discrete but has continuous parents (",
names(parents)[illegal.parents], ").")
# extract the CPTs.
cpt.levels = dimnames(new)
parents.levels = smartSapply(NULL, names(parents),
function(p) dimnames(parents[[p]]$prob)[[1]])
# check the number of dimensions.
if (length(cpt.levels) != length(parents) + 1)
stop("wrong number of parents for node ", node, ".")
# check the dimension names, and add them if they are missing
if (is.null(names(cpt.levels)))
names(dimnames(new)) = names(cpt.levels) = c(node, names(parents.levels))
else {
# check whether the CPT covers all the relevant variables.
if (!setequal(names(cpt.levels)[-1], names(parents.levels)))
stop("wrong dimensions for node ", node, ".")
# now that we are sure that the dimensions are the right ones, reorder
# them to follow the ordering of the parents in the network.
d = names(dimnames(new))
new = aperm(new, c(match(node, d), match(names(parents.levels), d)))
cpt.levels = dimnames(new)
}#ELSE
# check the levels in the CPT against those of the parents, and reoder them
# if needed.
for (p in names(parents))
if (!setequal(parents.levels[[p]], cpt.levels[[p]]))
stop("the levels of the parent ", p, " of node ", node, " do not match.")
if (!isTRUE(all.equal(parents.levels, cpt.levels[-1])))
new = ndsubset(new, parents.levels)
return(new)
}#CHECK.DNODE.RVALUE.VS.PARENTS
# check the consistency of an assignment to a bn.fit.gnode.
check.gnode.rvalue = function(x, node) {
components = c("coef", "fitted", "resid", "sd")
labels = c(fitted = "fitted values", resid = "residuals")
# custom list of components.
if (!is.list(x) || any(names(x) %!in% components))
stop("the conditional probability distribution for node ", node,
" must be a list with at least one of the following elements:",
paste0(" '", components, "'"), ".")
if (("coef" %!in% names(x)) || !any(c("sd", "resid") %in% names(x)))
stop("at least the regression coefficients and either the residuals or ",
"the residual standard deviation are required for node ", node, ".")
if (!is.null(x$coef))
if ((length(x$coef) == 0) || !is.real.vector(x$coef))
stop("coef must be a vector of numeric values, the ",
"regression coefficients for node ", node, " given its parents.")
# check that the coefficients are stored as numeric values, not integers.
x$coef[] = as.numeric(x$coef)
for (comp in c("fitted", "resid"))
if (!is.null(x[[comp]]))
if ((length(x[[comp]]) == 0) || !is.real.vector(x[[comp]]))
stop(comp, " must be a vector of numeric values, the ",
labels[comp], " for node ", node, " given its parents.")
# check the standard deviation of the residuals.
if (!is.null(x$sd)) {
if (!is.non.negative(x$sd))
stop("sd must be a non-negative scalar, the standard deviation of the ",
"residuals of node ", node, ".")
if (!is.null(x$resid)) {
adj.sd = cgsd(x$resid, p = length(x$coef))
if (!isTRUE(all.equal(x$sd, adj.sd, check.attributes = FALSE, tol = 0.0005)))
stop("the reported standard deviation of the residuals of node ", node,
" does not match the observed one.")
}#THEN
}#THEN
else if (!is.null(x$resid)) {
# compute it from the residuals if not provided, it must be defined.
x$sd = cgsd(x$resid, configs = x$configs, p = length(x$coef))
}#THEN
# make sure that the standard error is not an integer variable.
storage.mode(x$sd) = "double"
# one residual for each fitted value.
if (!is.null(x$resid) && !is.null(x$fitted))
if (length(x$resid) != length(x$fitted))
stop("the residuals and the fitted values of node ", node,
" have different lengths.")
return(x)
}#CHECK.GNODE.RVALUE
# check an assignment against the bn.fit.gnode it will be assigned to.
check.rvalue.vs.gnode = function(new, old) {
if (!is(old, "bn.fit.gnode"))
stop("both nodes should be continuous (Gaussian) nodes.")
# same number of coefficients.
if (length(new$coef) != length(old$coefficients))
stop("wrong number of coefficients for node ", old$node, ".")
# same regression coefficients.
if (!is.null(names(new$coef))) {
# if the new coefficients have labels, they must match.
check.nodes(names(new$coef), graph = names(old$coefficients),
min.nodes = length(names(old$coefficients)))
# reorder the coefficients to match.
new$coef = new$coef[names(old$coefficients)]
}#THEN
else {
# copy the names from the spec.
names(new$coef) = names(old$coefficients)
}#ELSE
# same number of standard errors (fixed to 1).
if (length(new$sd) != 1)
stop("wrong number of standard errors for node ", old$node, ".")
# same number of residuals.
if (!is.null(new$resid))
if (length(new$resid) != length(old$residuals))
stop("wrong number of residuals for node ", old$node, ".")
# same number of fitted values.
if (!is.null(new$fitted))
if (length(new$fitted) != length(old$fitted.values))
stop("wrong number of fitted values for node ", old$node, ".")
return(new)
}#CHECK.RVALUE.VS.GNODE
# check an assignment to a bn.fit.gnode using only the names of its parents.
check.gnode.rvalue.vs.parents = function(node, new, parents) {
# all parents of Gaussian nodes must be continuous nodes.
illegal.parents =
sapply(parents, class) %!in% c("bn.fit.gnode", "bn.fit.cgnode")
if (any(illegal.parents))
stop("node ", node, " is Gaussian but has discrete parents (",
names(parents)[illegal.parents], ").")
# add the intercept, which is obviously not among the parents of the node.
parent.names = c("(Intercept)", names(parents))
# same number of coefficients.
if (length(new$coef) != length(parent.names))
stop("wrong number of coefficients for node ", node, ".")
# same number of standard errors (fixed to 1).
if (length(new$sd) != 1)
stop("wrong number of standard errors for node ", node, ".")
# if the new coefficients have labels, they must match.
if (!is.null(names(new$coef))) {
if (!setequal(names(new$coef), parent.names))
stop("wrong regression coefficients for node ", node, " (",
paste(setdiff(parent.names, names(new$coef)), collapse = " "), ").")
new$coef = new$coef[parent.names]
}#THEN
else {
names(new$coef) = parent.names
}#ELSE
return(new)
}#CHECK.GNODE.RVALUE.VS.PARENTS
# check the consistency of an assignment to a bn.fit.cgnode.
check.cgnode.rvalue = function(x, node) {
components = c("coef", "fitted", "resid", "sd", "configs")
labels = c(fitted = "fitted values", resid = "residuals")
# custom list of components.
if (!is.list(x) || any(names(x) %!in% components))
stop("the conditional probability distribution for node ", node,
" must be a list with at least one of the following elements:",
paste0(" '", components, "'"), ".")
if (("coef" %!in% names(x)) || !any(c("sd", "resid") %in% names(x)))
stop("at least the regression coefficients and either the residuals or ",
"the residual standard deviation are required for node ", node, ".")
if (!is.null(x$coef))
if ((length(x$coef) == 0) || !is.real.vector(x$coef))
stop("coef must be a matrix of numeric values, the ",
"regression coefficients for node ", node, " given its parents.")
# check that the coefficients are stored as numeric values, not integers.
x$coef[] = as.numeric(x$coef)
for (comp in c("fitted", "resid"))
if (!is.null(x[[comp]]))
if ((length(x[[comp]]) == 0) || !is.real.vector(x[[comp]]))
stop(comp, " must be a vector of numeric values, the ",
labels[comp], " for node ", node, " given its parents.")
if (!is.null(x$configs)) {
if ((length(x$configs) == 0) || !is.factor(x$configs))
stop("the discrete parents' configurations for node ", node,
" must be a factor.")
if (nlevels(x$config) != ncol(x$coef))
stop("wrong number of discrete parents' configurations for node ", node, ".")
}#THEN
# check the standard deviations of the residuals.
if (!is.null(x$sd)) {
if (!is.nonnegative.vector(x$sd))
stop("sd must be non-negative, the standard deviations of the ",
"residuals of node ", node, ".")
if (length(x$sd) != ncol(x$coef))
stop("the dimensions of sd and coef do not match in node ", node, ".")
if (!is.null(x$resid) && !is.null(x$configs)) {
adj.sd = cgsd(x$resid, configs = x$configs, p = nrow(x$coef))
if (!isTRUE(all.equal(x$sd, adj.sd, check.attributes = FALSE, tol = 0.0005)))
stop("the reported standard deviations of the residuals of node ", node,
" does not match the observed one.")
}#THEN
}#THEN
else if (!is.null(x$resid)) {
# compute the standard error from the residuals if possible.
if (is.null(x$config))
stop("sd is missing, and parents' configurations are required to compute it.")
x$sd = cgsd(x$resid, configs = x$configs,
p = ifelse(is.matrix(x$coef), nrow(x$coef), length(x$coef)))
}#THEN
# make sure that the standard error is not an integer variable.
storage.mode(x$sd) = "double"
# one residual for each fitted value.
if (!is.null(x$resid) && !is.null(x$fitted))
if (length(x$resid) != length(x$fitted))
stop("the residuals and the fitted values of node ", node,
" have different lengths.")
# if any, one parent configuration for each fitted value and residual.
if (!is.null(x$config)) {
if (!is.null(x$resid) && (length(x$configs) != length(x$resid)))
stop("parents' configurations and residuals of node ", node,
" have different lengths.")
if (!is.null(x$fitted) && (length(x$configs) != length(x$fitted)))
stop("parents' configurations and fitted values of node ", node,
" have different lengths.")
}#THEN
return(x)
}#CHECK.CGNODE.RVALUE
# check an assignment against the bn.fit.cgnode it will be assigned to.
check.rvalue.vs.cgnode = function(new, old) {
if (!is(old, "bn.fit.cgnode"))
stop("both nodes should be continuous (conditional Gaussian) nodes.")
# right dimensions for the coefficients.
if (!is(new$coef, "matrix") ||
!identical(dim(new$coef), dim(old$coefficients)))
stop("the regression coefficients for node ", old$node, " must be ",
"in a matrix with one column for each discrete parent and one ",
"coefficient for each continuous parent.")
# right row names for the coefficients.
if (!is.null(rownames(new$coef))) {
# if the new coefficients have labels, they must match.
check.nodes(rownames(new$coef), graph = rownames(old$coefficients),
min.nodes = length(rownames(old$coefficients)))
# reorder the coefficients to match.
new$coef = new$coef[rownames(old$coefficients), , drop = FALSE]
}#THEN
else {
# copy the names from the spec.
rownames(new$coef) = rownames(old$coefficients)
}#ELSE
# right column names for the coefficients.
if (!is.null(colnames(new$coef))) {
# if the new coefficients have labels, they must match.
check.nodes(colnames(new$coef), graph = colnames(old$coefficients),
min.nodes = length(colnames(old$coefficients)))
# reorder the coefficients to match.
new$coef = new$coef[, colnames(old$coefficients), drop = FALSE]
}#THEN
else {
# copy the names from the spec.
colnames(new$coef) = colnames(old$coefficients)
}#ELSE
# same number of residuals.
if (!is.null(new$resid))
if (length(new$resid) != length(old$residuals))
stop("wrong number of residuals for node ", old$node, ".")
# same number of fitted values.
if (!is.null(new$fitted))
if (length(new$fitted) != length(old$fitted.values))
stop("wrong number of fitted values for node ", old$node, ".")
if (!is.null(new$dlevels)) {
old.dparents = names(old$dlevels)
new.dparents = names(new$dlevels)
# different sets of discrete parents.
if (!setequal(old.dparents, new.dparents))
stop("different discrete parents for node ", old$node, ".")
# identical sets of discrete parents, but in a different order; match
# configurations.
old.config = old$dlevels
new.config = new$dlevels
for (i in seq_along(old.config))
old.config[[i]] = paste(old.dparents[i], old.config[[i]], sep = ":")
for (i in seq_along(new.config))
new.config[[i]] = paste(new.dparents[i], new.config[[i]], sep = ":")
old.config = interaction(expand.grid(old.config))
new.config = interaction(expand.grid(new.config)[old.dparents])
map = match(old.config, new.config)
# reorder coefficients, standard errors and configurations.
new$sd = new$sd[map]
new$coef = new$coef[, map, drop = FALSE]
new$dlevels = new$dlevels[old.dparents]
colnames(new$coef) = names(new$sd) = names(old$sd)
}#THEN
return(new)
}#CHECK.RVALUE.VS.CGNODE
# check an assignment to a bn.fit.cgnode against the local distributions of its
# parents.
check.cgnode.rvalue.vs.parents = function(node, new, parents) {
# both continuous and discrete parents are legal, but treated differently.
discrete.id = sapply(parents, class) %in% c("bn.fit.dnode", "bn.fit.onode")
discrete.parents = names(parents)[discrete.id]
continuous.id = sapply(parents, class) %in% c("bn.fit.gnode", "bn.fit.cgnode")
continuous.parents = names(parents)[continuous.id]
parent.configurations =
prod(sapply(parents[discrete.parents], function(p) dim(p$prob)[1]))
correct.configs = as.character(seq(from = 0, to = parent.configurations - 1))
new$dparents = which(discrete.id)
new$gparents = which(continuous.id)
new$dlevels = lapply(parents[discrete.parents], function(p) dimnames(p$prob)[[1]])
# the node should have at least one discrete parent, or it would not be CLG.
if (length(discrete.parents) == 0)
stop("node ", node, " is a conditioanl Gaussian node but has no discrete parent.")
# check the dimensions of the regression coefficients matrix.
if (!isTRUE(all.equal(dim(new$coef), c(length(continuous.parents) + 1L,
parent.configurations))))
stop("wrong dimensions for the regression coefficients matrix of node ",
node, ".")
# check and fix the coefficients names.
if (!is.null(rownames(new$coef))) {
regression.coefficients = c("(Intercept)", continuous.parents)
if (!setequal(rownames(new$coef), regression.coefficients))
stop("wrong regression coefficients for node ", node, " (",
paste(setdiff(continuous.parents, rownames(new$coef)), collapse = " "), ").")
new$coef = new$coef[regression.coefficients, , drop = FALSE]
}#THEN
else {
rownames(new$coef) = c("(Intercept)", continuous.parents)
}#ELSE
# check and fix the column names of the coefficients matrix.
if (is.null(colnames(new$coef)))
colnames(new$coef) = correct.configs
else if (setequal(colnames(new$coef), correct.configs))
new$coef = new$coef[, correct.configs, drop = FALSE]
else {
# wrong names: warn and rename.
warning("remapping levels of the discrete parents configurations ",
"for node ", node, ".")
colnames(new$coef) = correct.configs
}#ELSE
# check the number of standard errors, and check their names.
if (length(new$sd) != parent.configurations)
stop("wrong number of standard errors for node ", node, ".")
if (is.null(names(new$sd)))
names(new$sd) = correct.configs
else if (setequal(names(new$sd), correct.configs))
new$sd = new$sd[correct.configs]
else {
# wrong names: warn and rename.
warning("remapping levels of the discrete parents configurations ",
"for node ", node, ".")
names(new$sd) = correct.configs
}#ELSE
# check the discrete parents configurations.
if (!is.null(new$configs)) {
actual.configs = levels(new$configs)
if (length(actual.configs) != parent.configurations)
stop("wrong number of discrete parents' configurations for node", node, ".")
if (any(actual.configs != correct.configs)) {
if (setequal(actual.configs, correct.configs))
new$configs = factor(new$configs, correct.configs)
else {
# wrong levels: warn and rename.
warning("remapping levels of the discrete parents configurations ",
"for node ", node, ".")
levels(new$configs) = correct.configs
}#ELSE
}#THEN
}#THEN
return(new)
}#CHECK.CGNODE.RVALUE.VS.PARENTS
# check discrete parents configurations in list form for all kinds of nodes.
check.discrete.parents.configuration = function(config, node,
ideal.only = FALSE) {
# check whether the configuration is there.
if (missing(config) || !is(config, "list"))
stop("the parents configuration must be a list with elements named after the parents of ", node$node, ".")
if (is(node, c("bn.fit.dnode", "bn.fit.onode")))
allowed.parents = node$parents
else if (is(node, c("bn.fit.cgnode")))
allowed.parents = node$parents[node$dparents]
# check the node labels from the list names.
check.nodes(names(config), graph = allowed.parents,
min.nodes = length(allowed.parents), max.nodes = length(allowed.parents))
# check that only one value is provided for each parent, to get a single
# configuration.
for (c in names(config))
if (length(config[[c]]) > 1)
stop("only one value allowed for parent '", c, "'.")
if (is(node, c("bn.fit.dnode", "bn.fit.onode")))
levels = dimnames(node$prob)
else
levels = node$dlevels
# check whether the provided values are valid values.
for (c in names(config)) {
if (is.factor(config[[c]]))
config[[c]] = as.character(config[[c]])
if (config[[c]] %!in% levels[[c]])
stop("the value of parent '", c, "' must be a valid level.")
}#FOR
return(config)
}#CHECK.DISCRETE.PARENTS.CONFIGURATION
Try the bnlearn package in your browser
Any scripts or data that you put into this service are public.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.