Provides the "enrich" method to enrich list-like R objects with new, relevant components. The current version has methods for enriching objects of class 'family', 'link-glm', 'lm', 'glm' and 'betareg'. The resulting objects preserve their class, so all methods associated with them still apply. The package also provides the 'enriched_glm' function that has the same interface as 'glm' but results in objects of class 'enriched_glm'. In addition to the usual components in a `glm` object, 'enriched_glm' objects carry an object-specific simulate method and functions to compute the scores, the observed and expected information matrix, the first-order bias, as well as model densities, probabilities, and quantiles at arbitrary parameter values. The package can also be used to produce customizable source code templates for the structured implementation of methods to compute new components and enrich arbitrary objects.
Methods to enrich list-like R objects with extra components
When I develop a piece of statistical methodology it is not uncommon
that it depends on functionality or quantities that a list-like core R
object object_x
of a certain class could have but doesn't. An
example is objects of class link-glm
which only provide up to the
the first derivatives of the link function. It is also not uncommon
that the functionality or quantities that I need are hard-coded in
another package in the R ecosystem. Prominent examples include
implementations of gradients of the log-likelihood and information
matrices for specific model classes.
In such cases, I either contact the developers and ask them to provide a generic which I can then use, or copy some of the code and adopt it to what I'm doing. Both options can be rather time-consuming, and particularly the latter is rarely bug-free.
I believe that
users and developers should have direct access to useful functionality and quantities in the R ecosystem, epsecially if these include implementations of complex statistical quantities
quantities and functionality that are specific to a list-like object_x
should be components of object_x
The above motivated me to develop the enrichwith R package that
allows object_x
to be enriched with components corresponding to the
option enrichment_option
through the following simple call
enrich(object_x, with = enrichment_option)
The call is inspired by Donald Knuth's literare programing paradigm.
The main objective of enrichwith is to allow users and developers to directly use the enrichment options that other developers have provided, through a clean interface, minimising the need to adopt source code of others.
The purpose of enrichwith is to provide:
useful enrichment options for core R objects, including objects of
class lm
, glm
, link-glm
and family
(see, for example,
?enrich.glm
)
methods for producing customisable source code templates for the
structured implementation of methods to compute new components (see
?enrichwith
and ?create_enrichwith_skeleton
)
generic methods for the easy enrichment of the object with those
components (see, for example, ?enrich
and
?get_enrichment_options
)
The vignettes illustrate the enrichwith functionality through comprehensive, step-by-step case studies. These also include illustrations from recent research of mine on methods for statistical learning and inference.
Get the development version from github with
# install.packages("devtools")devtools::install_github("ikosmidis/enrichwith")
link-glm
objectsObjects of class link-glm
have as components functions to compute
the link function (linkfun
), the inverse link function (linkinv
),
and the 1st derivative of the inverse link function (mu.eta
).
enrichwith comes with a built-in template with the methods for the
enrichment of link-glm
objects with the 2nd and 3rd derivatives of
the inverse link function.
The get_enrichment_options
method can be used to check what
enrichment options are available for objects of class link-glm
.
library("enrichwith")standard_link <- make.link("probit")class(standard_link)# [1] "link-glm"get_enrichment_options(standard_link)# -------# Option: d2mu.deta# Description: 2nd derivative of the inverse link function# component compute_function# 1 d2mu.deta compute_d2mu.deta# -------# Option: d3mu.deta# Description: 3rd derivative of the inverse link function# component compute_function# 1 d3mu.deta compute_d3mu.deta# -------# Option: inverse link derivatives# Description: 2nd and 3rd derivative of the inverse link function# component compute_function# 1 d2mu.deta compute_d2mu.deta# 2 d3mu.deta compute_d3mu.deta# -------# Option: all# Description: all available options# component compute_function# 1 d2mu.deta compute_d2mu.deta# 2 d3mu.deta compute_d3mu.deta
According to the result of get_enrichment_options
, the object
standard_link
can be enriched with the 2nd and 3rd derivative of the
inverse link function through the option "inverse link derivatives".
enriched_link <- enrich(standard_link, with = "inverse link derivatives")class(enriched_link)# [1] "enriched_link-glm" "link-glm"cat(format(enriched_link$d2mu.deta), sep = "\n")# function (eta)# {# -eta * pmax(dnorm(eta), .Machine$double.eps)# }cat(format(enriched_link$d3mu.deta), sep = "\n")# function (eta)# {# (eta^2 - 1) * pmax(dnorm(eta), .Machine$double.eps)# }
enriched_link
is now an "enriched" link-glm
object, which, as per
the enrichment options above, has the extra components d2mu.deta
and
d3mu.deta
, for the calculation of 2nd and 3rd derivatives of the
inverse link function with respect to eta
, respectively.
family
objectsbetareg
obejctsglm
and betareg
objectsget_information_function
for betareg
objects returns a function that can compute observed information matrices (by setting type = "observed"
)get_information_function
now returns a function that can compute the Cholesky decomposition of the information matrix (by setting CHOL = TRUE
)simulate
methods (coming form get_simulate_function
) now check for correct length of coefficient vector if this is specifiedenrich.family
can return 4th derivatives of the a function and other characteristics of the exponential familyenrich
method for betareg
objects (enrichment options with scores, information, bias and simulate)enrich.family
including a description of the exponential familywith
argument of enrich
methodsbias
function was passed a named vectorsget_dmodel_function
, get_pmodel_function
, get_qmodel_function
enriched_glm
can now be used to fit GLMs and get objects that are
enriched with auxiliary functions and other components when compared
to their glm
counterparts?enrich.family
get_dmodel_function.glm
,
get_pmodel_function.glm
, get_qmodel_function.glm
lm
objectscontributions = TRUE
for getting score contributionsglm
objectsglm
objectsglm
objectsglm
objects