The hardware and bandwidth for this mirror is donated by METANET, the Webhosting and Full Service-Cloud Provider.
If you wish to report a bug, or if you are interested in having us mirror your free-software or open-source project, please feel free to contact us at mirror[@]metanet.ch.
str2lang and
hcl.colors which require it, and added backport linter to
list of linters to identify whether the minimum version of R is required
to be changed in future builds.rmarkdown and igraph to
VignetteBuilder field of DESCRIPTION because these are
required to build the vignettes.abortif and
abortifnot to use base R rather then rlang.
Replaced legacy calls to rlang::abort in classes with calls to
abortif or abortifnot.LeafNode class,
and modified method QALY in LeafNode to
calculate QALYs gained, assuming present value of future utility under a
continuous discount assumption. Revised documentation of
DecisionTree and LeafNode to clarify the
timing assumptions used by rdecision in decision tree
modelling (that all costs are incurred at time t = 0, and that QALYs are
gained over the time horizons defined in the leaf nodes).SemiMarkovModel and DecisionTree
documentation.set_cost and set_utility
in class MarkovState without an argument is trapped as an
error.set_cost in class
Transition without an argument is trapped as an error.index and ref in the
tornado method of DecisionTree may now each be
expressed as a single Action edge, rather than a list of action edges,
if there is one decision node in the tree. As many decision problems are
a choice between two strategies, this is a common usage of
tornado, and will lead to more compact code. Method
is_strategy also modified to permit compact notation of
index and reference strategies.to make it available for creating such diagrams given any input source (e.g., Markov models), rather than restricting its use toModVar`s
in decision trees.SM00, a narrative explanation of
replicating Sonnenberg and Beck’s three-state model of people having a
prosthetic heart valve. Removed dependency of the package readme file on
test-model-Lifestyle.R and test-model-PHV.R, and revised, simplified and
corrected the readme file.stopifnot assertions to ensure that the
vignette model results agree with published values. Replaced calls to
knitr::kable with base R table output for purled output.stopifnot assertions to ensure that the
vignette model results agree with published values. Replaced calls to
knitr::kable with base R table output for purled output.stopifnot assertions to ensure that the
vignette model results agree with published values. Replaced calls to
knitr::kable with base R table output for purled output.stopifnot assertions to ensure that the
vignette model results agree with published values. Replaced calls to
knitr::kable with base R table output for purled output.stopifnot assertions in the vignette to ensure
that vignette model results agree with published values. Removed
test-model-SM02.R from testthat folder. Replaced calls to pander and
knitr::kable with base R table output so that tables and plots are
displayed when purled code is executed using source with echo.stopifnot assertions in the
vignette to ensure that vignette model results agree with published
values. Removed test-model-Tegaderm.R from testthat folder and added
extra tests to test-DecisionTree. Replaced calls to pander and
knitr::kable with base R table output so that tables and plots are
displayed when purled code is executed using source with echo.stopifnot assertions in the vignette to ensure
that vignette model results agree with published values. Removed
test-model-AZT.R from testthat folder. This is because some package
build and check environments break during vignette building if the
vignette R markdown refers to external scripts.gbp from vutils.R (in vignettes) to
utils.R, and removed vutils.R. This function is now available to the
package, and vignettes no longer need to refer to it in an external
file. Removed functions gv2png and placeholder
which are no longer required due to use of igraph for
drawing models.DiagrammeR and
added a suggested package dependency on igraph. Diagrams of
Markov models in package vignettes and readme file are now drawn with
package igraph. Extra code is required to produce desired
appearance, compared with using GraphViz, but the resultant image files
are compact and can be generated without the need for a command line
tool.expressions from
modvar_table in classes DecisionTree and
SemiMarkovModel. This is because operands of expressions
are included in the table, and it makes no sense to exclude expression
model variables which depend on those operands.modvars function in various classes to use
vectorization. Added to base classes Node and
Edge to avoid type checks.Graph.
This was causing errors under certain circumstances when creating graphs
with named lists, e.g.,
g <- Graph(V = list(a = nodea, b = nodeb), E = list(e = edgee))
because internal graph logic tests using identical gave
false negatives.as_gml to classes Graph and
Digraph for exporting of graphs in the Graph Modelling
Language (GML). Graph representations written in this format can be
imported to other packages, such as igraph to permit
modification of nodes, edges and their attributes.rankdir, width and
height to function as_DOT in
Graph for consistency with function of the same name in
Digraph. Removed requirement for all nodes to have labels
before using them by identifying nodes with their ID, and using the
label attribute in the file.set_utility to class
MarkovState to allow setting dynamically.get_cycle to class SemiMarkovModel
to query the current cycle number.get_tcycle to class SemiMarkovModel
to query the cycle duration.hcc.QALY to methods for cycling in class
SemiMarkovModel. The hcc arguments control
whether half cycle correction is applied to the values written to the
Markov trace but do not affect the value of the end-cycle population
passed to the next cycle, or when the state populations are queried. The
new argument permits half cycle correction to be applied independently
to the trace values of state populations, costs and QALYs, which can be
useful if, for example, it is desired to display end-cycle populations
with mid-cycle costs and utilities.draw of class DecisionTree
to avoid unintentional clipping of node labels. Modified algorithm to
make better use of unit conversions that are implicit in package grid
and to ensure that symbols for nodes retain the correct aspect ratio
when drawn in plot windows that are not square.evaluate in class
DecisionTree to check that the sum of probabilities for
each strategy is one, for each run. Confirmed that the check operates
correctly when the conditional probabilities of traversing an edge are
scalars or ModVars.1 - p where p is sampled from a Beta
distribution, for example. Revised the Tegaderm vignette to demonstrate
this feature.Action and Reaction to
disallow values of NA for cost and benefit.set_interval to LeafNode to allow
dynamic setting. Disallowed a utility of NA.LogNormModVar and
LogNormDistribution to reference the article in
BioInformatics.test_that expectations, and referenced code from the
vignette’s markdown. This avoided duplication of code between the
vignette and test code.set_utility to LeafNode.evaluate method of DecisionTree to
ensure that the components of strategy labels are ordered by the
lexicographical order of node labels. For example, if there is a
decision node labelled d1 with emanating action edges
d1a and d1b, and a decision node labelled
d2 with emanating action edges d2a and
d2b, the set of strategy labels is
{d1a_d2a, d1a_d2b, d1b_d2a, d1b_d2b}, not
{d2a_d1a, d2a_d1b, d2b_d1a, d2b_d1b}, even if node
d1 has a greater index than node d2.evaluate_walks method of
DecisionTree to return the index of the leaf node for each
path as the row name of the numeric matrix. Formerly it was coerced into
a real value and saved in the matrix in a column called
Leaf.reset function of SemiMarkovModel
to default to zero population in each state. Previously the first state
was allocated 1000 people, but if vertexes were reordered, the first
state is not the same as the first one provided, leading to unexpected
behaviour.postree method of Arborescence to
use lexicographic order of node labels to define sibling order. Changed
local array indexing to use [[]] and fixed unintentional
use of array slicing under some circumstances.par in the tornado plot section of
DecisionTree with calls to withr::with_par.
This avoids the need to change the global plot defaults. Package
withr added to dependency list.test_that expectations, and referenced code from the
vignette’s markdown. This reduced duplication of code between vignette
and test harness.vertex_label and
edge_label to Graph to support checked
iteration of lists of nodes and edges when retrieving labels.edge_index, edge_at
and has_edge in Graph.vertex_index,
vertex_at and has_vertex in
Graph.arrow_source and
arrow_target to Digraph to support checked
iteration of lists of arrows to retrieve their source or target
nodes.test-model-Sumatriptan.R), to avoid replicating
code in test scripts.set_cost, set_benefit and
set_probability to class Reaction to allow
dynamic setting of costs, benefits and probabilities in decision trees
without the need to rebuild the model.set_cost and set_benefit to class
Action to allow dynamic setting of costs and benefits in
decision trees without the need to rebuild the model.vutils.R in folder vignettes as a home for
helper functions used in vignette building.SemiMarkovModel by creating
private methods to manage the kernel operations of cycling the
population and tallying the costs and benefits. Approx factor of 10
improvement in speed, which is helpful with PSA.hcc.pop and hcc.cost
in function cycle of SemiMarkovModel. Removed
the requirement for hcc.pop to be TRUE if
hcc.cost is TRUE (i.e., the corrections are applied
independently).test-model-AZT.R), taking
code from the SM01-HIV vignette and the SemiMarkovModel test script.
Non-test chunks are referenced by the vignette to avoid repetition.
Edited and clarified the vignette and added PSA (as per Briggs example
4.7). Used DiagrammeR to create Markov model diagram.test_that context, thus avoiding repetition of code between
the vignette and the test case.rankdir, width and
height in Digraph::as_DOT to permit the
drawing direction and size of canvas to be adjusted in the dot
syntax.abortif
and abortifnot to be ..., to increase their
similarity with stopifnot, allowing testing of multiple
conditions with a single call.threshold in DecisionTree now has
no default value for parameter outcome.chance_nodes in DecisionTree now
has the option to return nodes, indices or labels, consistent with
similar functions for decision nodes and leaf nodes.Digraph now contains
integers.Graph and
Digraph now contain integers.abortif, which is identical to
abortifnot but with a negated condition. Intended to make
argument checks more readable by avoiding double negatives.is_probabilistic for ExprModVar
now returns correct result if the expression involves integers
(previously and erroneously it returned NA).is_probabilistic for base class
ModVar now returns FALSE rather then
NA.as_value to
as_numeric to more accurately reflect its behaviour.DecisionTree.MarkovState$utility.u <- ConstModVar$new(description = "", units = "", const = 2)
and passed as the utility argument to
LeafNode. Scalar arguments remain subject to range checking
in [-Inf,1] for normal usage and to retain the previous behaviour.as_value to return the value
of an object if it is a ModVar (via its get()
method) or if it is a numeric object. Intended as an internal function
to avoid type testing on polymorphic variables.is_ModVar to test whether an
object is a model variable. Intended as an internal function to the
package.abortifnot (a replacement for
stopifnot using rlang but with limited capability). It is a
non-exported function intended for use in checking function arguments
without increasing the cyclomatic complexity of the function
itself.W (list of
walks) for DecisionTree$evaluate_walks. Added the
alternative argument Wi in which the indices of edges in
each walk are provided, to improve efficiency in avoiding repeated
conversion between edges and indices during PSA. Added function
edge_properties to collect information for computing path
products and sums without repeated tree walking.vertex_along, edge_along,
vertex_at and edge_at to class
Graph to clarify the relationship between nodes and edges
and their indices, and to help optimise iterations in graph
algorithms.SemiMarkovModel$cycle() by
generating intermediate results as matrices.ExprModVar to
check that nested autocorrelation is supported (i.e. when at least one
model variable appears twice or more as an operand of an expression,
when it is evaluated recursively).set for
ModVar and ExprModVar in the documentation for
those classes.ExprModVar now has an empirical distribution,
which is sampled on creation, to optimize functions mu_hat,
sigma_hat and q_hat, at Paola’s
suggestion.EmpiricalDistribution and its test
harness.CohortMarkovModel to
SemiMarkovModel in README.OccCost and EntryCost columns in
SemiMarkovModel$cycle to make them per person costs.SemiMarkovModel.SemiMarkovModel as caution for
converting between transition rates and per-cycle probabilities. Cited
work of Jones et al
set_probabilities method to
SemiMarkovModel to set transition probabilities from a
matrix.DirichletDistribution class, mainly
to support PSA in semi-Markov models.BetaModVar has a BetaDistribution
uncertainty.ModVar with a “has-a” relationship to an
underlying uncertainty distribution. Incorporated ability to link
several model variables to a common underlying distribution (for use
with multinomial Dirichlet etc.).DiracDistribution.Distribution for each of the
currently supported distributions (Beta, Normal, Log Normal,
Gamma).Distribution to represent multivariate
distributions.SemitMarkovModel and its test script.Transition (inherits from
Node) and its test script.MarkovState (inherits from
Edge) and its test script.value to method set in class
ModVar. This allows variables to be set to an explicit
value; used in threshold finding.threshold function to DecisionTree
to calculate the value of a model variable at which the cost difference
becomes zero or ICER crosses a threshold.run to by argument of
DecisionTree$evaluate(). Avoids application having to
reshape output before reporting PSA results.DecisionTree$tornado which caused
bars to be clipped under some circumstances.testthat/setup.R.as_DOT to Graph and
Digraph for export to graphviz DOT file format to aid
visualization of graphs.tolerance in expect_equal with
abs in expect_true for approximate equality tests.WORDLIST file and sundry administrative changes
for clean package build.README file, with an example and
acknowledgements.draw() method to DecisionTree.tornado() method to DecisionTree for
univariate sensitivity analysis.DecisionTree (1000 evaluations of a typical HTA tree takes
< 5s on a typical PC).ModelVariable as ModVar for
compactness, and renamed its derived classes similarly.ModelVariable.ModelVariable.CITATION file to inst folder in CRAN
preparation.tests/testthat folder with tests for
ModelVariable.ModelVariable class as the new base
class from which to construct the variables in an economic model. The
class includes methods to support parametrization of uncertainty in the
model variable.ModelVariable to model
particular forms of uncertainty. These are
ConstModelVariable, NormalModelVariable,
GammaModelVariable, BetaModelVariable,
LogNormalModelVariable. They do as expected from their
names. Some support alternative forms of parametrization.ExpressionModelVariable. A sub-class of
ModelVariable, objects of this class are defined with an
expression involving other model variables. The concept permits
variables to be combined in any mathematical expression that R itself
will support. Because ExpressionModelVariables are
themselves ModelVariables, they can can appear in an
expression that is used to define another model variable.Node and its sub classes to accept
ModelVariables as arguments to costs, utilities and
probabilities, thus embedding probabilistic sensitivity analysis into
decision tree models.ModelVariable approach to PSA.Node classes.node.apply and path.apply
functions, and subsumed them into Node.node.apply and
path.apply, and subsumed them into Node.Node objects with a Document Object Model
(DOM) interface, as far as practicable.references.bib to directly embed them in the YAML headers.
To do: explore whether references can be saved in preferred bib
format.nullfile(), for suppressed output, in
function des with detection of OS to support older R
versions (nullfile was introduced to base R at 3.6.0).desNode,
LeafNode, ChanceNode,
DecisionNode) and pathway detection and traversal
functions.These binaries (installable software) and packages are in development.
They may not be fully stable and should be used with caution. We make no claims about them.