This PsN version has been tested on the following systems:

Windows 10 64-bit with Strawberry Perl 5.24.0 , gfortran 4.9.2 and NONMEM 7.1.0, 7.1.2, 7.2.0, 7.3.0 and 7.4.0
Windows 7 64-bit with Strawberry Perl 5.18.2, gfortran 4.7.3 and NONMEM 7.2.0 and 7.3.0
Windows 7 64-bit with Strawberry Perl 5.24.1, gfortran 4.9.2 and NONMEM 7.4.0
Windows 7 64-bit with ActivePerl 5.16.3 (32-bit), gfortran 4.6.0 and NONMEM 7.3.0
MacOS X 10.6.6 with ActiveState Perl 5.16.3, ifort 12.0.0 and NONMEM 7.2.0, 7.3.0 and 7.4.0
Ubuntu Linux 16.10 with Perl 5.22, gfortran 6.2 and NONMEM 7.2.0, 7.3.0 and 7.4.0
Scientific Linux 6.3 with Perl 5.10.1, gfortran 4.4.7 and NONMEM 7.2.0
Scientific Linux 6.3 with Perl 5.10.1, gfortran 5.1.1 and NONMEM 7.3.0 and 7.4.0
CentOS 7 with Perl 5.16.3, gfortran 4.8.5 and NONMEM 7.3.0 and 7.4.0


************************
Installation
************************

Follow the installation instructions on
https://uupharmacometrics.github.io/PsN/install.html

************************
Documentation
************************
All documentation is found in the doc subdirectory of the PsN 
installation directory. During installation it is also possible 
to have all documentation copied to a user-specified directory.

All documentation can also be downloaded from
the PsN website https://uupharmacometrics.github.io/PsN/docs.html

************************
Dependencies 
************************

- It is recommended to use Xpose version 4.5.3 or later for the -cwres functionality 
  and the default execute, vpc and bootstrap R plots.

- Each userguide lists the R packages required to obtain the diagnostic R plots.

************************
Known issues 
************************

- Windows users need to be careful to choose 
  a Perl version where the extra modules needed for PsN are available. 
  For new Perl installations the PsN team recommends Strawberry Perl
  See the instructions on https://uupharmacometrics.github.io/PsN/install.html

- Installation problem on some Windows systems where 
  perl 'system' command does not work. Make sure perl module File::Copy::Recursive
  is installed *before* (re)running the installation script setup.pl

*********************************************
Changes and bug fixes from PsN 4.6.0 to 4.7.0
*********************************************

A new recommended but not required Perl library for runrecord: YAML::Tiny 



   ___General___

- New web page at https://uupharmacometrics.github.io/PsN/
- New git repository and development webpage at https://github.com/UUPharmacometrics/PsN
- PsN has updated its license. It is still licensed under the GNU public license version 2, but
  it is now possible to relicense PsN under any later versions of this license.
- File specified in $ETAS will be automatically copied when model run
- Bug fix: Some generated paths for the Rplots script were illegal R strings causing R to crash
- Runs where any of the models have $NONPARAMETRIC will now get an additional file in the main
  run directory: raw_nonparametric_<modelname>.csv, containing regular ofv, nonparametric ofv, etc
- The handling of $MSFI and MSFO=file has been redesigned, also for crash restarts. MSF-files
  never have to be listed with -extra_files or -extra_output.
- Bug fix: PsN can now handle $MSFI including/path/to/file.msf
- Bug fix: PsN will notify the user if the equal sign in IGNORE=character is skipped,
  which is not supported by PsN and hence not correctly handled when reading datasets for e.g.
  bootstrap.
- Added support for IGNORE=" (double quotation mark) in $DATA
- Bug fix: If Perl fork() fails when running locally on unix, then sleep for 30s and try
  again. If keeps failing then treat model as failed due to job submission error, and
  allow crash restart if appropriate crash handling options are set.
- Bug fix: PsN can now be run from symbolic link wrappers.
- Added seconds to "Run started" in version_and_option_info.txt
- Added finish time to version_and_option_info.txt


   ___common_options___

- Option -abort_on_fail has a new meaning and function. The option is intended for the system tests.
  If set, PsN will exit with an error if NONMEM crashes, or an estimation/evaluation fails to
  produce an ofv, or a simulation gives a typical error message in the lst-file.
- Option -standardised_output was renamed to -so
- New option -rmarkdown. If -rplots>0, some tools by default will create an Rmarkdown file. 
  To create R file instead of Rmarkdown file, use -no-rmakdown.


   ___NONMEM output parsing___

- Added NM7.4 NONMEM output extensions .npl, .vpd and .clt to the list of recognized extensions.
- Bug fix: Handle error message "0ERROR IN READING FILE : FCON" in
  eigenvalue section of lst-file.
- Bug fix: If the execution of a $PROBLEM stops before any output is produced, consider the
  NONMEM run failed only if it is the first $PROBLEM. For example, if the first $PROBLEM is
  an estimation and the second a simulation, and the estimation completes but the simulation
  is not initiated due to e.g. poor estimation results, then PsN will no longer treat the run
  as crashed.
- Bug fix: PsN can now parse NONMEM output from runs with only nonparametric, no parametric,
  estimation/evaluation.
- Bug fix: PsN can now handle $MSFI in combination with NONMEM's sparse matrix format in lst-files,
  which is used for very large covariance matrices.
- Bug fix: Two digits of precision were lost when parsing OFV from .ext file
  

   ___Cluster/Grid___

- When -run_on_torque is set, PsN will now add -d <full path to NM_run subdirectory>
  to the qsub command. The qsub command will be saved to file qsubcommand in the
  NM_run subdirectory, for easier debugging.


   ___Test scripts___

- The temp directory for the unit and system tests is configurable in psn.conf. See
  Testing chapter in developers_guide.pdf.
- runsystem will redirect PsN log messages to a file in the user's home directory,
  or a directory set in psn.conf. The set of 'all' nm_versions in 'runsystem all'
  command is configurable in psn.conf. See Testing chapter in developers_guide.pdf.


   ___execute___

- New option -boxcox to automatically boxcox transform all ETAs before running the model
- New option -etas to automatically add eta starting values from the previous phi-file
- If -rplots>0, by default Rmarkdown file will be created.


   ___runrecord___

- Option -to is no longer required. The default value is the highest run number found in
  an uninterrupted sequence of files <root>1.<mod_ext>, <root>2.<mod_ext>,... etc.

New features implemented by William S. Denney of Human Predictions:
- Output to .csv using a sufficient part of RFC 4180 [https://tools.ietf.org/html/rfc4180]
  to handle commas and quotes (optionally, not by default).
- Optionally output to .csvy files (.csv files with a YAML header; readable by the rio R
  library).  Columns can now have attributes like parameter type
  ('theta', 'omega', ...). The YAML header requires the YAML::Tiny Perl library,
  if it is not installed the header will have text format.
- Automatically detect if the output file has the extension "csv" or "csvy" and output to
  the csv/csvy format if yes and another output format has not been specifically requested.
- Automatically detect max_lvl (optionally, by default)
- Give a header column name to the child run headers to more easily distinguish them from
  empty column headers (which now should not exist) (optionally, by default).


   ___sumo___

- Bug fix: The message stating that there is no gradient information after an SAEM run was duplicated
- The checks for large relative standard errors were removed together with the options to set the limits.
- A new message when the NONMEM run failed, i.e. estimations/evaluation failed to produce an ofv,
  or when simulation had special error message in lst-file.
- Do not print message about rounding errors or parameter near boundary if the NONMEM run failed.
- Bug fix: When the model file copy at the top of the lst-file has $MSFI and there are no parameter
  labels to use, PsN will use the estimation results to guess which parameters have been estimated
  and list them in sumo output using their generic names.


   ___bootstrap___

- The $COVARIANCE will by default be removed from the model. If wanted it can be kept
  with the new -keep_covariance option.
- Print a warning message if the number of individuals in the data set is different from the number
  of individuals NONMEM printed in the original run lst-file.
- New file 'all_individuals1.csv' contains a list of all individuals of the original dataset.
- If -rplots>0, by default Rmarkdown file will be created.


   ___vpc/npc___

- New option -mix to generate simulated and observed data for external post processing
- Refactored input checking.
- Fix bug causing warnings to be emitted when having '.' in DV column of original table.
  The warnings came from the calculation of the diagnostic False Pos. It is now defined that
  such observations can neither be false positives nor false negatives.
- If -rplots>0, by default Rmarkdown file will be created.


   ___frem___

-  will, if a covariate-covariate correlation is exactly 0, set intial omega estimate corresponding to
   1% correlation in Model 2. Similarly in Model3, if correlation from phi-file exactly 0, or
   parameter-parameter covariance has to be added to create full block, then set initial estimate as 1% correlation
-  handle PRED-models without EVID/MDV column, i.e. all data records are observations.
-  handle model 2 phi file ETA all 0
-  new option -mu, i.e. use mu-modelling for added covariates. Not set by default.
-  new option -estimate_means to true. When option -estimate_means is set, only covariates which have
   missing observations in the frem dataset will have the mean estimated. If -no-estimate_means is set,
   no means will be estimated regardless of missing observations.
-  rearranged control stream changes. Now all frem code is put last in $PRED/$ERROR and no new variables
   are defined (this is needed for very large numbers of covariates).
-  New option -estimate_means, default not set. If set then estimate covariate means instead of setting
   FIX to mean from observed data,
-  New option -mceta, default not set. Set MCETA=N in last $ESTIMATION.
-  New option -deriv2_nocommon_maxeta. Default 60. Set $ABBREVIATED DERIV2=NOCOMMON if number of ETAs
   is larger than this number.
-  Put intermediate models in subdir intermediate_models instead of in m1.
-  Handle PHI in addition to ETA in phi-file (non-classical methods), and multiple $EST (only
   results from last est are used).
-  set NONINFETA=1 in last estimation if it is a classical method and NM-version > 7.2.
-  New postprocessing script postfrem 
-  Rescale option is now default on.
-  Create approximate proposal density for sir even if covariance step of Model 4 is successful.
-  If the covariate covariance matrix is non-positive definite after rounding
   then modify it to make it positive definite before using it as initial
   estimate for the new OMEGA block in Model 2.
-  If frem is resumed, for example after manual modification of the initial
   estimates in a model and removal of the corresponding lst-file, then
   frem will automatically recreate all control streams that depend on the
   modified model and rerun them.


   ___scm___

- Disallow restarts in existing directory as this operation did not work
- scm will treat . (dot) in covariate column of data set as missing value 
- bug fix option time-varying: handle PRED models without MDV in $INPUT
- Change the default lower bounds for the THETA in power and exponential
  parameterizations from -100000 to -100.
- Removed special use of common option -abort_on_fail.
- Bugfix: If -model was given on command line but not in config file PsN would crash.
- If -rplots>0, by default Rmarkdown file will be created.

   ___linearize___
- bug fix option time-varying: handle PRED models without MDV in $INPUT
- If the phi-file of the original model is available linearize will use that in a
  $ETAS record.
- New option -keep_covariance


   ___boot_scm___

- Disallow restarts in existing directory as this operation did not work


   ___xv_scm___

- Disallow restarts in existing directory as this operation did not work


   ___sse___ 

- New option -update_fix, to be used in combination with -rawres_input. Default false.
  Update initial estimates for parameters that are FIX.
- bug fix when samples=1. RSE should be NA and not 0.


   ___simeval___

- Bug fix: Skip computations on residuals if CWRES is zero.
- Handle ID:s without observations, sporadic zero-valued EBE:s
- New rplots
- If -rplots>0, by default Rmarkdown file will be created.
- Possibility to rerun the post processing without rerunning NONMEM by using -dir
- New option -dv for DV vs PRED plot


   ___update_inits___

- New options -rawres_input, -in_filter, -offset_rawres to read parameter estimates from a raw results file.
- New option -etas to add $ETAS and MCETA=1 to the model file using the phi-file
- Bug fix: renumber MSFO=file in $NONPARAMETRIC


   ___parallel_retries___

- New option -update_fix, default false. Only for use in combination with -rawres_input.
  Update parameters that are FIXED.


   ___sir___
	
- new option -cap_correlation, default 0.8. If set to a number between 0 and 1, at every iteration ensure
  that no absolute value of a correlation in the proposal exceeds this number.
- message suggesting reparameterization if any correlations exceed 0.8 (before or after capping)
- updated R template
- changed inflation principle, covariances are not inflated meaning that correlations are in practice decreased 


   ___pvar___

- Add OFV and run number columns to results.csv


   ___cdd___

- New option -update_inits, set by default. The initial estimates in the cdd models will be
  the final estimates from the base model, unless -no-update_inits is set. If -no-update_inits
  is set, then the initial estimates will be as in earlier versions, i.e. the same as
  the initial estimates of the base model.
- New option -etas to use $ETAS and the phi-file of the original model for all estimation runs
- changed directory name prefix for cross-validation sub-directory (if
  option -xv is set) from modelfit to crossval.
- ID is now default for the case_column option
- New rplots added.
- If -rplots>0, by default Rmarkdown file will be created.


 ___rplots___

- Major redesign


   ___npfit___

- New tool npfit for nonparametric estimation, see npfit_userguide.pdf


   ___transform___

- New tool for automatic model transformations.


   ___crossval___ 

- New option -msf, to use MSFO-MSFI instead of updating initial estimates in
  prediction models. New format of output file xv_results.csv, renamed from xv_results.txt.
- Automatic nonparametric cross-validation if input model has $NONPARAMETRIC.  
- Refactoring of cross-validation code.
- A new userguide crossval\_userguide.pdf