Changes to clinDR:
======================
Update 2.4.1
- The prior distribution for fitEmaxB and emaxsimB by default now
truncate the ED50/P50 distribution to (0.001,1000), and truncate
the lambda distribution to (0.3,4.0). Along with adjustments to
the MCMC default control parameters, divergences reported by STAN
have been largely eliminated. These changes, along with some changes
to the iternal parameterization of the STAN model, produce
small numerical changes from past versions of clinDR
Update 2.3.7
- Function fitEmaxB can now be applied to first-stage fitted
output with correlation between the initial dose group estimates.
The function checkMonoEmax is now deprecated and replaced by the
function bpchkMonoEmax with input now the output object of
function fitEmaxB.
- The shiny app has been renamed runSimulations. It has been
extensively updated
Update 2.3.6
- Added print function for prior objects created by
emaxPrior.control(). The output is a text file that
describes the prior distribution. There is an option ('sap')
for a more detailed version appropriate for an SAP, and a
shorter version for protocols. The RandEmax function is
now deprecated and replaced by randomEmax, which generates
data with Emax model parameters now generated from a
distribution described by the object output of
emaxPrior.control().
Update 2.3.5
- Added shiny interface for emaxsim(B) initiated with runShiny()
- Changed the default prior scale parameter for loglambda from
0.85 to 0.425. This change effects fitEmaxB and emaxsimB
- Changed the default sample size in mcmc.control to yield 9999
samples of the model parameters by default
Update 2.3
- Added percentiles of the posterior distribution of the
model parameter to the output object of emaxsimB
- Standardized default Emax model to 4-parameter sigmoid.
Changes made to fitEmax and emaxsim that may affect
execution of older code.
- Standardized confidence/posterior level to clev=0.9.
Changes made to summary.emaxsimB, plotB, plotD, targetCI.
- SE calculation adjusted by add 1/2 0/1 to rates in
emaxsim(B) to avoid NA in pairwise comparisons with
low response rate data
- Several minor bug fixes in emaxsim
Update 2.2
- metaData updated with removal of a duplicate
compound-indication. Meta-prior default values
also updated in second decimal places.
Update 2.1
- New parameterization of the Emax function has been implemented in
fitEmaxB and emaxsimB. The 'emax' parameter has been replaced by
the difference in response at a targeted dose versus placebo. Corresponding
changes have been made in the supporting methods. Each parameter nows
has a prior t-distribution with DF that can be specified. The default
values in the prior distributions are based on meta-data updated through
August 2019.
- The proportion of divergent MCMC iterations for each emaxsimB
simulated analysis is now returned as part of the output
- Plot symbols are printed on top of error bars for better visibility
- Several bug fixes for function plotB relating to plots with active
controls. xat option added to control axis labeling. The dac option was
removed.
- Plotting of Bayes estimates in fitEmaxB and plotB is now based (default) on the posterior median
rather than the mean to avoid problems with occassion extreme values in MCMC output. Posterior
medians were added to means in output of predict.fitEmaxB.
- Changed log option to logScale option in plotting functions to avoid masking of the log function.
- Meta data in examples14 and examples16 has been replaced by the data metaData, which aggregates all
of the dose response meta data from previous studies through August 2019. The meta data from all
compounds are now in a single file with a common format.
- Added functions solveDose and solveEmax. solveDose returns the dose yielding a target
response for the input Emax model parameters. solveEmax returns the Emax parameter
yielding a target response at a specified dose
Update 1.9.5
- Additional input error checking
- pboAdj in fitEmax can now fit multiple protocols with/without placebo
adjustment
= Changed check for centered covariates in fitEmax(B) to accept larger difference with zero
- Created new metaData with all dose response studies included in
the meta-analyses through Aug 2019. A new unified data format is used.
The older examples14 and examples16 data have been removed (they are
part of metaData)
Update 1.9
- Add support for baseline covariates in fitEmax, fitEmaxB and supporting
functions
- Add showStanModels function to display the STAN code for Emax models
- Corrected output of popSD naming in emaxsim extraction function
- Consolidated Bayesian Emax models into a single STAN model file.
Specification of model type (e.g., hyperbolic vs sigmoidal) is no
longer required in selEstan and showStanModels functions
- The 'logit' parameter in function checkMonoEmax was re-named 'binary'
- Return printed matrices as invisible objects in print.emaxsim(B) and
print.emaxsim(B)obj
Update 1.8
- Added testMods of class Mods from R package DoseFinding to the
output object of emaxsim(B). testMods gives the functions defining the
contrasts for MCP-MOD testing. The ED50contr and lambdacontr that
can also be used to define the contrasts were removed.
- Added coef,sigma, and vcov methods for fitEmax(B), emaxsim(B) objects
- The summary.emaxsim and summary.emaxsimB functions now
return the printed output for extraction. Summaries for
the best pairwise comparison were added.
- Corrected use of specified starting values in function emaxalt and emaxsim.
- Improved error handling in emaxsim(B) to reset random number generator
Update 1.7
- The calling sequence for 'plotPop' in plot.emaxsimobj and
plot.emaxsimBobj have been changed to ensure a valid value
is input.
- The parameter difPlacebo has been renamed plotDif in
plot.emaxsim and plot.emaxsimB for consistency with other
plotting functions
- Corrected bug in MCP-MOD test with negative trends
- Corrected bug causing warning messages from DRDensityPlot
when the quantiles have row or column names
- Corrected plotting of population curve at placebo in
plot.emaxsimB
Update 1.6
- Change package dependencies
- Additional documentation for CRAN release
- Modify examples for CRAN release
Update 1.05
- Fix bug in parallel processing for emaxsim and emaxsimB
- Fix bug in residSD input to customCode in emaxsimB
Update 1.04
- RandEmax can now generate random residual SD from a scaled inverted chi distribution
Update 1.03
- Added prediction interval option to plot.emaxsimobj
- Expanded output from emaxsimB when check=TRUE so
test data for custom code can be generated
Update 1.02
- Improved labeling of output of predict.fitEmax(B)
- Add descriptive label to plot.emaxsim(B) to indicate difference with placebo by default
- Corrected bug in emaxsim output of residual SD. This impacts predictive intervals from several functions.
Update 1.01
- Major update adding Bayesian versions of fitEmax and emaxsim
- Supporting methods, e.g., plot, print,... also supplied for
new Bayesian functions fitEmaxB and emaxsimB
- Parallel computing options added for the Bayesian functions
Update 0.9.6.5
- Added log scale option to plot.plotB.R, plotD.R, plot.fitEmax.R, plot.emaxsimobj.R.
If plot=plotDif (difference with placebo) and log=TRUE,
the plot starts from the lowest nonzero dose.
- Improved active control plotting in plot.plotB.R.
- Added optional user-specified control over x-axis tick marks.
Update 0.9.6.4
- In checkMonoEmax, included equality in predictive probability calculation.
- In plot.fitEmax, if there is no grouping variable, then no panel at the top
of the plot is created. The function now returns a list with the ggplot2
object and a matrix with the confidence and prediction interval limits.
Update 0.9.6.3
- Added grouped-data option to checkMonoEmax, and test data.
- Corrected plotting issue in plotB.
Update 0.9.6.2
- Updated code for fitEmax test statistics. plot.fitEmax predictive
error bars now use the smallest sample size for a dose/strata combination.
- plotB now returns the ggplot object so it can be further modified.
- fitEmax can now use non-integer prot variable.
Update 0.9.6
- Added binary logistic Emax model fitting to emaxsim and all associated
functions
- Changed check of the variance-covariance matrix in fitEmax to ensure it
is positive definite
- Eliminated femax and femaxH
- Replaced RandEmax with skewRandEmax and eliminated the name skewRandEmax
Update 0.9.5.1
- Corrected bug in predictive interval for binary data in plot.fitEmax
Update 0.9.5
- Added treatment difference option to plot.fitEmax
- Updated implementation of symbol group in plotB
- Corrected bugs in prediction intervals in plot.fitEmax
- Added treatment difference estimation to predict.fitEmax
- Updated emaxsim and its support functions to use fitemax
and its support functions
Update 0.9.4
- Added GOF tests to fitEmax output
- Fixed symbol bug in plot.fitEmax and plot.logitEmax
- Fixed fitEmax and logitEmax to return NULL when convergence fails.
- Created fitEmax to unify fitting of Emax models with continuous/binary data
Update 0.9.3
- Updated all graphic output to use ggplot2
- Changed order of Emax model parameters to better support fitting
of data from multiple protocols with different placebo response.
Order of parameters in logitEmax also changed for consistency across
all Emax functions.
- Created new function contEmax to fit continuous emax models using
ML estimation along with supporting plot and predict functions
- Added plot symbol options and residual plots to plot.contEmax and
plot.logitEmax
- Fixed bug in 4-parameter binary data case for startEmax
- Fixed bug in predictive intervals in plot.logitEmax
- Changed default confidence/posterior interval levels to 0.90
Update 0.9.2
- led50 replaces ed50 everywhere.
- Changed options for skewRandEmax
- Last version before changes to graphics.
Update 0.9.1
- Numerous changes to DoseResponse package. The biggest change was to eliminate
transformation of the Hill parameter.