Breaking Changes
Parametric Models
rank_regression(): For all distributions, the
confidence intervals of the parameters are now computed on the basis of
a heteroscedasticity-consistent (HC) covariance matrix. If the
confidence intervals for the Weibull parameters are to be calculated
according to Mock, this must be specified via the new argument
options.
mixmod_regression(): Since this function uses
rank_regression(), the changes regarding the confidence
intervals also apply here.plot_prob.wt_model(): Removed dysfunctional argument
distribution. The distribution is inferred using the model
x.
Confidence Intervals
delta_method(): Renamed argument p with
x.confint_betabinom() and confint_fisher():
Removed constant features distribution, bounds
and direction from the tibble output and added them as
attributes instead.
Monte Carlo Simulation
mcs_mileage(): Changed name of output column
mileage to x (in accordance with
reliability_data()).mcs_delay(): Changed name of output column
time to x (in accordance with
reliability_data()).dist_mileage.default() (former
dist_mileage()): Renamed argument mileage with
x.mcs_mileage.default() (former
mcs_mileage()): Renamed argument mileage with
x.
New Features
Distributions
- Implementation of one- and two-parametric exponential distribution
(
'exponential' and 'exponential2').
Non-Parametric Failure
Probabilities
estimate_cdf(): Added option
johnson_method to specify the formula which is used for
determining cumulative failure probabilities.
Parametric Models
rank_regression(): New arguments direction
(specifies direction of dependency in the model), control
(enables access to argument control in
optim()) and options (method used to calculate
the confidence intervals for the parameters, default is “HC”).
r_squared_profiling(): New argument
direction.ml_estimation(): New arguments
start_dist_params (optional vector with initial values of
the parameters) and control (enables access to argument
control in optim()).loglik_profiling(): New argument wts.loglik_profiling() is now an S3 generic.
loglik_profiling() becomes
loglik_profiling.default(). Added
loglik_profiling.wt_reliability_data().loglik_function() is now an S3 generic.
loglik_function() becomes
loglik_function.default(). Added
loglik_function.wt_reliability_data().
Confidence Intervals
confint_betabinom(): Methods "kaplan" and
"nelson" of estimate_cdf() can be used for
beta-binomial confidence bounds.
Monte Carlo Simulation
- Added
mcs_mileage_data(): Create consistent MCS data
for mcs_mileage().
- Added
mcs_delay_data(): Create consistent MCS data for
mcs_delay().
dist_mileage() is now an S3 generic.
dist_mileage() becomes dist_mileage.default().
Added dist_mileage.wt_mcs_mileage_data().dist_delay() is now an S3 generic.
dist_delay() becomes dist_delay.default().
Added dist_delay.wt_mcs_delay_data().dist_delay() now supports the estimation of multiple
delay distributions at once.mcs_mileage() is now an S3 generic.
mcs_mileage() becomes mcs_mileage.default().
Added mcs_mileage.wt_mcs_mileage_data().mcs_delay() is now an S3 generic.
mcs_delay() becomes mcs_delay.default(). Added
mcs_delay.wt_mcs_delay_data().- Added
print.wt_mcs_delay_data() and
print.wt_mcs_mileage_data().
- Added
print.wt_mileage_estimation().
- Added
print.wt_delay_estimation() for one delay and
print.wt_delay_estimation_list() for multiple delays.
Lifecycle changes
Minor Improvements and bug
fixes
Reliability Data
- Fixed bug in
reliability_data(): Using !!
syntax with arguments x and status resulted in
an error.
estimate_cdf() preserves additional columns, that were
returned from reliability_data(..., .keep_all = TRUE).- Improved
print.wt_reliability_data().
Confidence Intervals
- Fixed bug in
plot_conf(): Wrong confidence bounds were
displayed for direction = "x" (#181).
- Fixed bug in
plot_conf():
plot_method = "ggplot2" and exactly one method in
estimate_cdf() resulted in an error (#182).
Monte Carlo Simulation
- The object returned by
mcs_mileage() now has class
wt_mcs_mileage.
- The object returned by
mcs_delay() now has class
wt_mcs_delay.
- The object returned by
dist_mileage() now has class
wt_mileage_estimation.
- The object returned by
dist_delay() now has class
wt_delay_estimation or
wt_delay_estimation_list.
Documentation improvements
plot_prob(): Better work out the distinction between
plot_prob.wt_cdf_estimation() and
plot_prob.wt_model(). The former is applied to a CDF
estimation whereas the latter is applied to a mixture model.
Breaking Changes
- Package now depends on R(>= 3.5.0)
Non-Parametric Failure
Probabilities
mr_method(): Deprecated, use
estimate_cdf() instead. Renamed output column
characteristic with x. Set default value for
id to NULL.johnson_method(): Deprecated, use
estimate_cdf() instead. Renamed output column
characteristic with x. Set default value for
id to NULL.kaplan_method(): Deprecated, use
estimate_cdf() instead. Renamed output column
characteristic with x. Set default value for
id to NULL.nelson_method(): Deprecated, use
estimate_cdf() instead. Renamed output column
characteristic with x. Set default value for
id to NULL.plot_prob.default() (former plot_prob()):
Renamed event with status.plot_prob_mix(): Deprecated, use
plot_prob() instead. Removed default value
NULL for argument mix_output. Renamed
event with status.
Parametric Models
ml_estimation.default() (former
ml_estimation()): Renamed event with
status. Removed details. Changed names and
contents of list elements in output. See
?ml_estimation.loglik_function: Renamed event with
status. Renamed pars with
dist_params.rank_regression.default() (former
rank_regression()): Renamed event with
status. Removed details. Changed names and
contents of list elements in output. See
?rank_regression.mixmod_em.default() (former mixmod_em()):
Renamed event with status. Removed
post.mixmod_regression.default() (former
mixmod_regression()): Renamed event with
status. Added arguments k and
control, which provide finer control over the segmentation
process. Expect default setting to provide other results than in prior
versions.predict_prob(): Renamed loc_sc_params with
dist_params.predict_quantile(): Renamed loc_sc_params
with dist_params.plot_mod.default() (former plot_mod()):
Renamed event with status. Renamed
loc_sc_params with dist_params. Removed
y.plot_mod_mix(): Deprecated, use plot_mod()
instead. Renamed event with status.plot_pop(): Added argument tol to restrict
the range of failure probabilities. Removed argument color.
Renamed argument params to dist_params_tbl,
which only supports location and scale parameters (also for
distribution = "weibull"). Changed behavior of
dist_params_tbl: A tibble is now recommended
instead of a vector.
Confidence Intervals
confint_betabinom.default() (former
confint_betabinom()): Renamed event with
status. Renamed loc_sc_params with
dist_params. Added argument b_lives which
allows the user to specify probabilities p for
B_p-lives to be considered.confint_fisher.default() (former
confint_fisher()): Renamed event with
status. Renamed loc_sc_params with
dist_params. Renamed loc_sc_varcov with
dist_varcov. Added argument b_lives which
allows the user to specify probabilities p for
B_p-lives to be considered.delta_method(): Renamed loc_sc_params with
dist_params. Renamed loc_sc_varcov with
dist_varcov.plot_conf.default() (former plot_conf()):
Switched position of arguments direction and
distribution.
Monte Carlo Simulation
dist_delay_register(): Deprecated, use
dist_delay() instead.dist_delay_report(): Deprecated, use
dist_delay() instead.mcs_delay_register(): Deprecated, use
mcs_delay() instead. Renamed x with
time. Renamed event with status.
Removed seed. Removed int_seed from output
list.mcs_delay_report(): Deprecated, use
mcs_delay() instead. Renamed x with
time. Renamed event with status.
Removed seed. Removed int_seed from output
list.mcs_delays(): Deprecated, use mcs_delay()
instead. Renamed x with time. Renamed
event with status. Removed seed.
Removed int_seed from output list.dist_mileage(): Removed event. Renamed
x with time. Switched position of arguments
time and mileage.mcs_mileage(): Removed event. Renamed
x with time. Switched position of arguments
time and mileage.
New Features
- Added support for ggplot2 in all plot functions. Plot method can be
selected in
plot_prob() or plot_pop() via
argument plot_method.
- Added
reliability_data(): Create consistent reliability
data.
- Added
estimate_cdf(): Unite functionality of
mr_method(), johnson_method(),
kaplan_method() and nelson_method(). Added
option ties.method for method = "mr", which
specifies how ties should be treated.
- Support of multiple methods in
estimate_cdf() and all
functions that depend on the cdf_estimation
(rank_regression(), plot_prob(),
plot_mod(), plot_conf(),
mixmod_regression()).
plot_prob() and plot_mod() are able to
handle mixture models.mixmod_regression() is now more flexible. Argument
k can be used to control number of subgroups or to
determine them in an automatic fashion. Argument control
provides additional control over the segmentation procedure.- Added
print.wt_rank_regression(),
print.wt_ml_estimation(),
print.wt_model_estimation(),
print.wt_model_estimation_list(),
print.wt_mixmod_regression() and
print.wt_mixmod_regression_list().
- Added
vcov.wt_model_estimation().
- Added
dist_delay(): Generalizes the
distribution-specific modeling of delays.
- Added
mcs_delay(): Generalizes the adjustment of
operating times by delays and supports multiple delays at once.
- Added lifecycle badges
Introduction of S3 interface
rank_regression() is now an S3 generic.
rank_regression() becomes
rank_regression.default(). Added
rank_regression.wt_cdf_estimation().plot_prob() is now an S3 generic.
plot_prob() becomes plot_prob.default(). Added
plot_prob.wt_cdf_estimation() and
plot_prob.wt_model().plot_mod() is now an S3 generic.
plot_mod() becomes plot_mod.default(). Added
plot_mod.wt_model().plot_conf() is now an S3 generic.
plot_conf() becomes plot_conf.default(). Added
plot_conf.wt_confint().plot_pop(): Added support for multiple population lines
and comparison of two- and three-parametric distributions.
Documentation improvements
- Revised README.
- Revised vignettes.
- Capitalized parameter documentation.
Lifecycle changes
Deprecated
dist_delay_register() and
dist_delay_report(): Use dist_delay()
instead.mcs_delay_register(), mcs_delay_report()
and mcs_delays(): Use mcs_delay()
instead.mr_method(), johnson_method(),
kaplan_method() and nelson_method(): Use
estimate_cdf() instead.plot_prob_mix(): Use plot_prob()
instead.plot_mod_mix(): Use plot_mod()
instead.
Removed
calculate_ranks.mixture_em_cpp.plot_layout.
Minor improvements and bug
fixes
- Fixed bug inside
plot_mod_mix() for the case of no
mixture distribution.
- Fixed bug inside
confint_betabinom(): many cases near
one -> unique().
- Fixed bug inside
mr_method(): assigning a rank for the
same lifetime.
- Fixed bug inside
mixmod_regression: call to
segmented::segmented.lm() was incorrect.
- Added trace type
"scatter" and scatter mode
"markers" to plotly plots.
delta_method(), r_squared_profiling() and
loglik_profiling() were vectorized.
- Fixed installation error when using clang compiler
Prerequisite for Package
Usage:
- Since RcppArmadillo is used, the R version should be at least 3.3.0
(listed under Depends in DESCRIPTION file)
Changes
- Vignettes for non-parametric probability estimation, parameter
estimation using Median-Rank Regression and Maximum-Likelihood and
mixture model estimation are provided.
- Argument y in functions
plot_prob_mix() and
plot_mod_mix() is deprecated and not used anymore.
- Argument reg_output in functions
plot_prob_mix() and plot_mod_mix() is
deprecated; use mix_output instead.
- Function
plot_mod_mix() was revised and updated in the
way that the obtained results of the function mixmod_em()
can be visualized.
- Function
plot_prob_mix() was revised and updated in the
way that the obtained results of the function mixmod_em()
can be visualized.
- Implementation of EM-Algorithm using Newton-Raphson. The algorithm
is written in c++ (
mixture_em_cpp()) and is called in
mixmod_em().
- New method for the computation of Fisher’s Confidence Bounds
regarding probabilities is used. These method is called “z-Procedure”
and is more appropriate to manage the bend-back behavior. Therefore an
adjustment of functions
delta_method() and
confint_fisher() was made.
- Implementation of log-location-scale models with threshold parameter
like three-parametric Weibull (“weibull3”), three-parametric lognormal
(“lognormal3”) and three-parametric loglogistic (“loglogistic3”).
- Implementation of location-scale models like smallest extreme value
(“sev”), normal (“normal”) and logistic (“logistic”).
- Implementation of Log-Likelihood Profiling for
three-parametric models in function
loglik_profiling(). In
general this function is used inside ml_estimation() for
the purpose of estimating threshold parameter of three-parametric
models.
- Implementation of R-Squared Profiling for three-parametric
models in function
r_squared_profiling(). In general this
function is used inside rank_regression() for the purpose
of estimating threshold parameter of three-parametric models.
- Implementation of Log-Likelihood Function for all
implemented models in function
loglik_function(). In
general this function is used inside ml_estimation() for
the purpose of estimating the variance-covariance matrix of
location-scale models “sev”, “normal” and “logistic”. The function is
also used to estimate the variance-covariance matrix of
log-location-scale models with a threshold parameter, i.e. “weibull3”,
“lognormal3” and “loglogistic3”.
- new argument in function
ml_estimation():
wts for case weights.