Restored npde function by creating new git npde repository
None
None
Fixed error in makePost associated with conversion to tidyverse handling
New function update_gfortran()
A flag error using iFort, thanks @sabryr.
Fixed a bug when rendering the html report.
the css is backed up at the lapk.ork server, now users can see the report /w styles also if they dont have pmetrics installed (remote execution)
Fixed bug in makePost that would cause failure if outeq>1
Supporting R 4.0.0
A bug that made some tables in the report have wrong headers was fixed.
Pmetrics will register each new installation.
Now the local version number is compared with the version number in github after suggest to update Pmetrics.
Updated plot.PMvalid with error checking for icen and outEq.
PMupdate is deprecated.
Fix makeValid and plot.PMvalid to handle multiple outEq.
Depecrated functions were exported.
This version is a pre-release beta for version 2.0
makeValid and plot.PMvalid added to replace makeNPDE and add pcVPC capacity
NPDE code folded into Pmetrics since original NPDE package no longer maintained
Windows users will now see terminal window open automatically
Added remote server run capabilities, currently only on LAPKB server
Restructured start up and installation to be self contained for all modules, including gfortran
Cleaned printing of summary.PMfinal object
Upated package NAMESPACE
None
MMopt now has multi-weighting option for any combo of auc, max, min
makeValid and plot.PMvalid added to replace makeNPDE and add pcVPC capacity
Updated makePTA to accept PMpost and PMmatrix objects
PMtree now checks for existing project and aborts if found to prevent overwriting.
Fixed minor bug in npc to make calculated percentage between 5th and 95th percentile correct
Updated for R version 3.5
None
Added the removal of lines with only spaces in model files that could cause an error
Clarified instruction line to launch batch file in Windows to execute NPAG/IT2B run.
Fixed bug in SIMrun when trying to used pre-loaded data file
Fixed bug with copying priors from previous run now that runs made inside new folders
make_AUC now returns NA when only one observation present for a subject
Models with more than 9 compartments detect compartments correctly
Restored split function to simulator, which was broken in 1.5.0
Recoded PMcompare to comply with deprecated recycling of vector length 1 in R, as of v. 3.4.3
Updated PMwrk2csv to properly handle files with 0 rates, i.e. only bolus dosing
NPAG can now use variables which are fixed in the population, but estimated
Added ability to use external dataset for NPDE
Added shrinkage calculation to final objects as final.x$shrinkage. Also added this to the summary table in PMreport.
makeNCA has new terminal argument (default=3) to control number of concentrations used for fitting k to terminal portion of time-concentration curve.
Added weighting to MMopt.
Removed dependency on rJava for PMcheck
Added ability to simulate with both covariates and split. The mean covariate value is added to each support point in this case.
makePTA with target.type = 'time' now MUCH faster thanks to improvements by Jan Strojil
Added the ability to makePTA to recalculate success threshold and redefine labels using existing PMpta object
SIMparse can now operate in parallel and will check if older simulations are present.
Updated Shiny code for PMcode(run)
The mult argument works properly in plot.PMsim now.
Fixed bug in makePTA and simulated targets that miscalculated success
Fixed bug in makeNPDE that caused a crash when model and data files have same basename.
Corrected makeFinal postPoints when nsub>100.
Improved handling of missing subjects in makeNCA.
Added a check to PMcheck for subjects with no observations
Added binSize argument to plot.PMsim to bin simulated concentrations, which is useful for VPCs.
Fixed plot, summary, and print PMop to handle objects with no observations
PMtest updated for parallelized version of Pmetrics
fixed SIMrun ID bug
NPAG code parallelized over points introduced. Default is serial for algebraic/exact and parallel for ODE models. Can be overridden by user.
Post-run report now includes Processing mode (parallel | serial) and run time.
Added the ability to specify predInt argument to SIMrun as a list of multiple intervals to generate predictions.
When making a csv file from simulated output, missing obsNoise now defaults to model file values, not 0.
Added PMmb2csv to convert old .mb or USC*PACK files to .csv format
Added setPMoptions and getPMoptions for session options. Currently they are sep (datafile field separator which can be "," or ";") and dec (decimal separator which can be "." or ","). Removed PMreadDefaults and PMwriteDefaults as Defaults package no longer supported.
Bug with ADDL and II in NM2PM fixed
Allowed makeFinal to work with older NPdata objects that don't have postden items
Changed incorrectly formatted values <1E-100 to be 0 in output files
Fixed makeFinal and makePost to handle one subject data files properly
Fixed PMwrk2csv when no assay error coefficients present
General updating of help files, which isn't really a bug!
Fixed bug in simulator output when values very near 0 didn't format properly
None
Fixed makeFinal bug regarding posterior density
Scaled bivariate NPAG final plots properly
Added Bayesian posterior parameter points to final object as final$postden
Added a plot method to plot.PMfinal to plot posterior distributions for a given parameter and set of subjects.
covariate argument now works properly for alphanumeric ID and when no covariates are fixed
Added summary.PMfinal to provide non-parametric parameter summaries based on simulation or parametric parameter summaries based on central limit theorem
Added col.pred argument to plot.PMmatrix to allow separate coloring of predicted profiles
Added makePTAtarget to simulate distributions of targets and updated makePTA, plot.PMpta, and summary.PMpta accordingly.
Added the ability to do PDI (pharmacodynamic index) plots with plot.PMpta.
NPAG, IT2B and simulator updated logic for SS dosing to have observation at time x be the concentration x hours after the beginning of the last steady state dose
Added another argument to the covariate SIMrun list argument: fix, which permits some covariates to be fixed to values in the template and not simulated
Re-parameterized limits argument to the covariate SIMrun list argument so that it is a named list with the desired limits for each covariate
Fixed bug in model file syntax checking with compilers other than gfortran.
summary.PMcov now handles non-numeric IDs correctly.
makePTA now calculates AUC correctly when start!=0 - it was adding in a time=0, conc=0 row, regardless of whether start=0 or not
Added time after dose plotting option to plot.PMmatrix.
Added summary.PMmatrix and print.summary.PMmatrix
SIMrun now checks for covariance matrices that are not pos-def and offers several solutions
Fixed bug in plot.PMmatrix where subjects with all missing observations caused an error.
add=T now functions properly in plot.PMsim when nsim<10.
Fixed error in makeNPDE causing problem with old instruction files.
Updated PMrun to work when data and model files have same root name
None yet
makeNPDE now works properly when the run to use was made with a non-uniform prior
SIMrun developed a bug after introducing covariate simulations such that fixed parameter values weren't handled properly. This has been corrected.
PMcheck had a bug when TIME=0 observation requirement removed. Fixed.
PMcheck now makes a color-coded .xlsx file with highlighted errors. This code was co-developed with Patrick Nolain in Limoges, France.
More widespread checking occurs for packages with automatic downloading/installation for those that are missing
makeNCA now takes a variety of arguments including a run number which can be used to do NCA on an data file or a post object, or a previously loaded data file, or a data file in the working directory
The post and pop objects after an NPAG run now contain the predictions at observation times in addition to the intervals specified by the idelta argument to NPrun
Model file format updated to TSMULTM
Removed PMcheck requirement for no observations at time=0. Predictions will always be 0 unless non-zero initial conditions or steady-state dosing specified.
Fixed bug in covariate simulations when 0-variance covariates are not present
Corrected plot.PMmatrix to handle error bars on observations when C0,C1,C2,C3 missing from data file
Included gamma/lambda in SD for weighting in bias/imprecision - now matches option 6/7 results in old USC*PACK
PMstep now works properly with a datafile containing only one covariate
AIC calc changed in IT2B/NPAG from -TRULOG + (P + Q) to 2*(-TRULOG + (P + Q))
Added covariate argument to SIMrun to enable simulation with covariates
Internal moving of parseBlocks function to independent function
Fixed small bug in makeNCA that caused an error in Tmax when there were two successive simulated Cmax that were the same by rounding.
Corrected formulae for weighted prediction error to weight by SD not variance
You can now plot any subset of output equations in plot.PMmatrix.
Fixed bugs related to making post and pop objects
Just bug fixes from 1.2.3
Numerous little bugs with make family and PMstep all related to adding mean and median predictions.
Updated to reshape2
Enhanced algorithm for makePost and makePop to generate data frames with predictions based on means and medians
Updated make_AUC, makeNCA, plot.PMmatrix to include icen (mean or median) argument to use new output from makePost and makePop
Updated makeOP and makeCov to eliminate icen argument and make objects that include predictions based on means and medians
Updated plot.PMop to use icen and pred.type (post or pop) rather than type
Updated plot.PMcov to use icen
Added PMsave as complement to PMload
Fixed bug where prior DENxxxx file was moved rather than copied from earlier run when using the prior=run notation in NPrun
makeCov now handles multiple input models properly
SIMparse and makePTA now handle missing values in simulated files
Fixed PMupdate by setting up package repository on LAPK website.
Added boolean option to suppress legend in makeErrorPoly
Added summary.PMcov to summarize covariates and Bayesian posterior parameter values for each subject
Added free.fraction argument to makePTA, as suggested by Joe Kuti
PMload can now load any number of results in a single command
Corrected time above target calculations in makePTA to reflect proportion of stated time interval, i.e. end-start
Added MMopt function for MM-optimal design from NPAG output
Added silent data file check when using NPrun, ITrun, or SIMrun.
Added error trap if the number of lines of output error coefficients in the model file is not equal to the number of output equations.
Added RMSE and %RMSE to summary.PMop and pop%RMSE and post%RMSE to PMcompare
Added checks for EVID and TIME values in all rows to PMcheck
Added check for non-numeric columns to PMcheck
Removed requirement to have all subjects end with an EVID=0 observation event
Reversed axes so that x is now predicted and y is observed to be consistent with mathematical notation obs = pred + error
Corrected handling of multiple output equations by PMcompare
Fixed creation of .config folder as needed on Mac/Linux systems
Stabilized name of directory for calling post-run processing script which couldn't be found by some Windows users
Corrected bug in simulator when multiplying limits with fixed variables present
Pmetrics now works with Linux! This is in prep for server mode and parallelized code.
Added NONMEM to Pmetrics translator: NM2PM. It has been tested on several NONMEM files, but will probably not handle every possibility yet.
Cleaned up NPrun, ITrun, ERRrun code to use common source which is a hidden function. This makes these functions more stable for changes moving forward.
Examples updated.
Documentation for PMcheck updated.
By entering nsim=0, you can now simulate one profile from each point in a non-parametric prior. This is prep for including the MMopt optimal sampling algorithm into Pmetrics.
A prior for NPrun can now be specified by simply typing the run number in addition to the existing methods of NPdata.x or DEN0001 file.
PMbuild will now correctly detect missing packages and install them.
Fixed logic in PMreport to better handle error conditions
Redesigned report and consolidated NPreport and ITreport into single PMreport function
Added tex file report to output (code by Alona Kryschenko)
Changed default behavior of obsNoise argument in SIMrun. Now, if missing, will default to 0 for all coefficients/output equations (i.e. no noise). To use values in data file first, and if missing, the model file, specify obsNoise=NA.
Fixed a bug in simulator related to parsing multiple output equations
None
Corrected error in instructions causing a crash when fitting data with no covariates
Added a check in NPrun, ITrun, ERRrun and SIMrun to ensure that model and data file names are 8 characters or fewer
Fixed more simulator bugs relating to dataframe format with lots of observations that continued on separate lines
Updated help for make_AUC to include method argument (linear or other)
Corrected handling of exclude argument to NPrun, ITrun and ERRrun
Change simulator argument to clean rather than cleanUp to be consistent with other functions
Simplified start messaging
Fixed simulator bug with limits related to new dataframe structure of simulator output
Corrected bug in makePTA when using a single target with new dataframe structure of simulator output
Fixed directory handling in Windows version of PMupdate.
Updated NPAG and IT2B to run off single instruction file rather than control stream plus instruction file
Added p-values to PMcompare based on nearest neighbors analysis of joint density for similarity. Also eliminated need for list of NPdata or ITdata objects.
Changed simulator output to a simpler dataframe structure
Fixed the detection of output equation numbers in model file building to account for conditional outputs
Set compartment amts from SIMparse to NA if a 0 compartment model (explicit output)
Fixed bug in PMcheck which did not properly detect a PMmatrix object supplied as the object to check.
Added exclude to list of arguments for SIMrun
Replaced PMdiag with makeNPDE which is a wrapper for npde::autonpde (Comets et al)
Added a fourth option to specify limits in SIMrun: a vector of 1 or 2 numbers to multiply the columns of the final$ab object used in the simulation.
Updated for Windows 8 and Mac OSX Mavericks
Internal changes to structure of PMsim objects from array to data.frame. Also added compartment amounts to PMsim list obtained from SIMparse.
Fixed a bug in PMupdate
Added ability to use wildcards with makePTA to parse Simulator files on the fly
Combined NPload and ITload into single PMload function
Added ability to include comment lines prefixed with 'C ' in model files
Added include and exclude to make_AUC function
Changed predInt argument to SIMrun to allow for start, stop, interval times.
Reworked SIMparse to allow for combining across simulations with different times, as well as to include or exclude files returned when using wildcards.
Added weighted.t.test to summary of PMop objects to test for bias different than 0. Code for this is courtesy of FelixS (http://www.r-bloggers.com/weighted-t-test-in-r/), also inspired by a function from Gavin Simpson on R-Help
Fixed incorrect handling of salt fractions for multiple drugs.
When outeq is specified with plot.PMsim and obs is included (for VPC), correct observations via outeq are now selected
Fixed bug that crept into PMcompare with plot=T option
Updated PMtree to generate template script with PMload rather than NPload
Added ability to use old .ZMQ files to NPAG
Now extracts Bayesian posterior densities for first 100 subjects into NPdata
Fixed notification to run PMbuild()
NEW! Pmetrics gets a GUI via the Shiny package. Use PMcode("command").
Moved the directory for compiled Pmetrics fortran code to the folder pointed to by Sys.getenv('APPDATA') on Windows systems
Added out argument to all Pmetrics plots enabling output to any graphic device, e.g. pdf, eps, tiff, jpeg, etc, making it easy to generate publication-ready plots.
Replaced PMcheckMatrix and PMcheckModel with PMcheck to check data files and (optionally) model files for errors.
PMcheck will now look for non-monotonically increasing times within a subject ID and subject IDs which have been reused, i.e. are non-contiguous.
The errbar argument is added to plot.PMatrix to plot standard deviation error bars around observations, based on C0, C1, C2, C3 in the data file or user-supplied values.
Exclude now added as argument to plot.PMmatrix, plot.PMop, and plot.PMcov to exclude subject IDs, i.e. the complement to the include argument.
PMfixMatrix deprecated and fix=F/T argument added to PMcheck
Fixed a bug that caused model files to parse incorrectly if headers included with no information. Now they are ignored.
Corrected handling of more than one fixed parameter when writing instruction files.
PMcheck (formerly PMcheckMatrix) now correctly reports subjects with first record != EVID 1 or TIME 0.
Minor housekeeping issues with some of the plot functions.
MAJOR update: the model file has been extended to eliminate the need for an instruction file by adding parameter ranges to PRI block and an ERRor block.
Added a new function to print errors in .csv files from PMcheckMatrix in a more easily digestible format.
Added a new function PMfixMatrix to attempt to automatically correct errors detected with PMcheckMatrix.
Added an 'extra' block to model file to allow for additional fortran subroutines to be added to the end of a model file
Changed makeOP and associated functions to always make OP objects for population and posterior predictions for all output equations, similar to the object loaded by NPload or ITload. This combined object also has class PMop. Functions can handle the larger object or the individual objects.
Reversed axes so that y is now predicted and x is observed to be consistent with mathematical notation that dependent variable is plotted on the y axis.
Redefined bias to be pred - obs, so that positive bias implies overprediction.
Changed output files to now be sequentially numbered rather than date-time stamped. This allows easier loading and examples.
Implemented system to compile Pmetrics objects upon first use, so that they do not have to be recompiled until there is a change in Fortran code.
Fixed bugs in SIMrun that mishandled multiple distributions manually specified and creation of .csv files
Corrected calculations for bias (mean weighted obs-pred) and imprecision (bias-adjusted mean weighted squared obs-pred)
BIG CHANGE: Model file is now a much simpler text file which is parsed to make the fortran code. Maintained backwards compatibility with existing fortran model files.
Added a growth dataset and qgrowth function to abstract CDC growth percentiles useful for simulating pediatric datasets.
Added bias and precision to observed vs. predicted regression plots
Clarified prediction errors on residual plots
Added bias and precision to PMcompare
Changed 'nsub' parameter to 'include' in SIMrun, so that non-contiguous subjects in a dataset may be used as simulation templates. This is analogous to 'include' in plot.PMmatrix.
Added 'predInt' option to SIMrun to automatically generate simulated predictions for each template at the specified prediction interval, including predictions at times which are explicitly included in the template (i.e. EVID=0 events).
By default, makeCov now does not summarize covariate values, but returns all measured values and the time of measurement. Bayesian posterior parameter values are replicated as necessary for each time/covariate entry.
plot.PMcov now accommodates the changes to makeCov above, so that when time is included as the x variable in a formula, one can plot changes in covariates or parameters over time, aggregated by subject. Plotting two parameters or a parameter vs. covariate results in the customary behavior as before.
PMstep is also updated to handle the new PMcov format.
Added include argument to plot.PMcov, plot.PMop
Users must now specify a data file in NPrun, ITrun and ERRrun. This enables more sophisticated error trapping. It also enables return of the data file as part of NPload or ITload.
Added checking for specified model, data, and (optional) instruction files and option to correct mistakes without aborting run.
Fixed bug in PMcheckMatrix when looking for missing covariate data in first entry for a patient
Fixed bug that caused crash when running population of 1 in NPAG
Fixed a bug in make_AUC for PMsim objects where the start parameter was ignored
Corrected plot.PMmatrix colors when single outeq > 1 plotted with prediction
Fixed another bug in plot.PMmatrix which did not handle block assignments correctly
Corrected subject IDs in PMcov objects
Corrected block (defined by EVID=4) handling in makeOP, make_AUC, makeNCA.
NPAG will no longer crash with a Hessian Matrix error, but will end 'nicely'
Changed 'denfile' argument in NPrun to 'prior' and enabled specification of either an NPdata object or a density file, e.g. DEN0001, for a non-uniform prior
Updated LAPK website with OS-specific instructions to download and install gfortran
Added intern=T option to NPrun, ITrun and ERRrun to run all three internally in R without a batch script.
Added PMstep function to carry out multiple linear regressions of covariates on Bayesian posterior parameters, with forward, backward, or forward/backward elimination of covariates
Users can now customize the defaults for any argument in any Pmetrics function using the setDefaults command from the Defaults package. A new Pmetrics command, PMwriteDefaults, will make these session defaults permanent until undone by unsetDefaults. See Defaults documentation for more details.
Corrected cycle numbering in plots when using a non-uniform prior for NPAG
Fixed bug in PMcheckMatrix if only one covariate
Corrected makeNCA to handle IV doses properly
PMtree() added to create basic folder substructure and skeleton R script for new Pmetrics project
PMupdate() added to update Pmetrics from within R without launching internet browser
Fixed a bug that crept into plot.PMsim which scrambled the simulated outputs
Error checking to NPreport() and ITreport() allowing creation/saving of partial data in case of error
Added PMnews() to simplify retrieval of Pmetrics changelog
Fixed bug in makePop with multiple outputs
Added makePop to generate predictions at frequent intervals for each subject using population parameter means, medians, or modes
Updated plot.PMmatrix, replacing "post" argument with "pred" to allow for either a post or pop object made by makePost or makePop, respectively
Added startup messages to inform users of news related to Pmetrics
Enabled plot.PMsim to plot one simulated profile correctly.
Added a warning to PMcheckMatrix if event times will exceed the maximum number of AUC intervals (48) in NPAG, based on the default interval of 24 hours.
Updated PMwrk2csv to handle new format with ADDL and II columns.
Added progress bars to parsing at the end of an NPAG or IT2B run
SIMparse can now open and combine several simulator output files at once.
Pmetrics now automatically checks for necessary fortran components and offers to download and install if missing.
Added a new command PMmanual() to launch the User's Manual and function catalogue.
Made makePTA() stable to simulated values that are missing or not real numbers
Corrected plot.PMsim to handle circumstances better when zeros are encountered for log plots
Fixed some bugs in SIMrun related to manually specified population priors
Fixed a bug in PMwriteMatrix that used improper line endings in Windows, causing extra blank lines to be written
Cleaned up bivariate plots in plot.PMfinal function.
Added target attainment function, summary and plot
Fixed bug in simulator which caused it to crash when only one random variable simulated
Fixed minor bug in plot.PMcycle which did not allow full customization of gamma/lambda plot
Added ADDL and II items to data matrix to support additional and steady state dosing. Old format matrix files without ADDL and II will be automatically updated. This change is propagated through NPAG, IT2B and the Simulator
Added the ability to specify ranges to truncate simulator output and avoid extreme, i.e. non-phyisologic, parameters
Added progress bar to PMdiag and cleaned display for SIMrun
Fixed bug in PMmatrix plot that did not permit customization of X-axis label
Corrected a bug in makeErrorPoly that prevented reporting of 3rd degree polynomial coefficients
Fixed handling of .csv file creation during a simulation using SIMrun with multiple subjects as templates
Fixed bug in ERRrun that prevented completion on Windows machines
Corrected the PMcheckMatrix function to properly look for missing covariate data at time=0 for each subject
Suppressed confidence intervals in plot.PMsim if nsim <100 and enabled it to handle missing confidence intervals properly
Cov objects are now loaded by default after NPAG and IT2B runs.
Added non-compartmental analysis to NPAG objects
Fixed matrix orientation of population means in simulator with multi-modal distributions
plot.PMcycle now works correctly if there are fewer than 10 cycles
Stopped makeErrorPoly from trying to open new graphics window
Updated makeErrorPoly to handle situations with too few points to fit 2nd and/or 3rd degree polynomials
Fixed errors in reading instruction files in ERRrun
Corrected mistake in IT2B runs with instruction files that caused end of file error
Took out AUC extraction with makeCov as it was too subject to error. This function is better implemented in the existing make_AUC command.
Pmetrics will now check for updates silently when it is first loaded into R with the library(Pmetrics) or require(Pmetrics) commands. If an update is found, the user will be notified with a simple, unobtrusive message.
The scripts will now handle security features associated with Windows Vista and later which restrict installation of packages into certain folders. Pmetrics now scans for the installation folder and sets this as an environment variable (which can be retrieved by Sys.getenv("PmetricsPath") for the remainder of the session. This is invisible to the user.
32 and 64 bit versions are now included.
None
Changed to compiled IT2B, NPAG, and Simulator objects. Initial roll out to support gfortran. More compilers added as needed by users.
Added an include argument to plot.PMmatrix to only plot a subset of the population. This can be used to match the same subset specified in an instruction file (which will be the only one modeled out of all subjects in the .csv file) or to simply focus on a subset of the total modeled population.
Fixed improper handling of block calculation by plot.PMmatrix and NPpostPred
Fixed bug in ITrun script that tried to use the tee logging command in Windows, which does not exist
Fixed another bug in ITrun which did not write correctly formatted integer to extnum file for run numbering purposes
Added override flag to PMwriteMatrix to write .csv file even with errors. Default is false. This now allows a .csv file to be written by PMwrk2csv even if there are errors. The errors will be reported, however.
Fixed improper handling of .csv files with comments by PMreadMatrix
Fixed improper handling of 0 cycle runs
Corrected path in Windows version to Rscript on some machines
Corrected logic in checking data files for missing covariate data at time 0 when there is only one covariate
Restored ability to control line characteristics in PMmatrix plot
Normalized .Library path to account for path truncation in Windows.
Fixed cycle plot of gamma from IT2B run to allow for one gamma for each output equation
Added npde, standardized visual predictive check with PMdiag command and object, which has a print method as well
Corrected plotting of output equations >1 in PMmatrix objects
Changed .libpaths() to .Library in NPrun, ITrun, ERRrun, SIMrun to avoid multiple library paths
Added additional items to ITparse and NPparse: tests for convergence, reporting of assay error model used, prior density used
Added support for different delimiters in PMreadMatrix
Added checking for column order and missing covariate data at time=0 to PMcheckMatrix
Fixed a bug in plot.PMmatrix that suppressed posterior plots from simulated data
Made Rscript work in NPrun and ITrun even if R not in user environment
Summary and print methods for PMop objects
Plotting summary of PMop as weighted residual vs. pred and vs. time
Greater control over grids in plots
Formula parsing in plot PMfinal
Column name checking in PMcheckMatrix
Regenerating seed file for simulations before each subject
Fixed axis ticks in log plots
So it begins.