This PsN version has been tested on the following systems: - MS Windows 7 with ActiveState 5.16.6 and NM 7.2.0. - Mac OS X 10.6.6 with Perl 5.10.0 and NM 7.2.0. - Linux Ubuntu 12.04 with Perl 5.14.2 and NM 7.2.0. - Linux CentOS 5.9 with Perl 5.8.8 and NM 6.1, 6.2, 7.1.0, 7.1.2, 7.2.0 and 7.3.0 beta. ************************ Installation ************************ - Depending on your Perl distribution you may have to install the following perl packages: Math::Random Storable::Store (Unix only) - Recommended perl packages, not absolutely required when running PsN but necessary for some features: Statistics::Distributions Archive::Zip File::Copy::Recursive All available from CPAN ( www.cpan.org ) - Before installing PsN, verify that all NONMEM installations you intend to use can be run directly via the nmfe (or NMQual) scripts and that they produce complete output files. Please note that with NMQual8 the nm72.xml file must be slightly modified (to expect .mod instead of .ctl as the control stream suffix) before NONMEM installation, see document psn_configuration.pdf, otherwise PsN's NMQual8 support will not work. This version of PsN has not been tested with NMQual, but is believed to work with NMQual version 8.2.4 and on. However the PsN reference output that comes with NMQual may be out of date. If you plan to use the parallelization features of NONMEM7.2 you must first verify that you can run NONMEM in parallel directly with nmfe72. If running NONMEM with nmfe (or NMQual) does not work then PsN will not work. - Unpack the installation package downloaded from psn.sf.net. It will create a directory called PsN-Source. Run the installation script setup.pl from within PsN-Source. - More detailed installation instructions are found on the PsN website http://psn.sourceforge.net/ under Installation. - The installation script can automatically generate a minimal configuration file psn.conf, both on Windows and Unix-type systems. If PsN is to be run on a cluster and/or with NMQual then some manual editing of psn.conf is still needed, see psn_configuration.pdf ************************ 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 http://psn.sourceforge.net/ ************************ Dependencies ************************ - Xpose must, if used, be version 4.2.1 or later to handle PsN-3.5.3 or later output format. It is recommended to use the latest Xpose release from http://xpose.sf.net ************************ Known issues ************************ - 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 - On Windows 7 with ActiveState Perl 5.16.3 the installation script failed to detect a correctly installed Math::Random module. If you think you have Math::Random installed even though the installation script complains then just go ahead with the installation and test running execute with option -min_retries=1 . If it works then you do have Math::Random and all is well. ********************************************* Changes and bug fixes from PsN 3.5.3 to 3.6.2 ********************************************* ___NONMEM interaction___ - A completely redesigned NONMEM output file parser. It has been tested on NONMEM versions 6.1, 6.2, 7.1.0, 7.1.2, 7.2.0 and 7.3.0 beta. NM5 has not been tested. The old parser in 3.5.3 is guaranteed to fail with NM 7.3 output file format. The new parser requires that the lst-file has a copy of the model file (control stream) at the beginning. This is handled automatically by NONMEM's nmfe scripts from version 6.2 and on. With NM6.1 you might have to edit nmfe6 to prepend the model file to the lst-file. PsN will exit with an error if the model file copy is missing. PsN uses the information in the model file copy to deduce which sections to expect in the rest of the lst-file, and e.g. to compare initial estimates with final estimates in the sumo program. - Support for $LEVEL $ANNEAL $ETAS $PHIS, new in NM 7.3. - PsN will remove $FINEDATA from the model file. - PsN requires that either option -nmfe or -nmqual is set, meaning that NONMEM must be invoked either via an nmfe script, an NMQual script, or a user written wrapper script. Most users do this already. It is no longer possible to make PsN do compiling and linking to produce the nonmem executable. - Running on slurm: More robust pattern matching in slurm queueing system responses. slurm_project is no longer a required parameter. - Running on torque: Improved monitoring of jobs, more robust pattern matching in torque system responses. - Bug fix: Handling of +Infinity and -Infinity strings in NONMEM output. - Bug fix: Handling of $AES and $AESINITIAL. PsN-3.5.3 could interpret and format these records incorrectly. - Bug fix: Successful covariance step will now be recognized by PsN. PsN-3.5.3 would sometimes fail at this. - Bug fix: Handling of estimation results from multiple $PROB in same model file. - Bug fix: More robust reading of date string in lst files(start and stop times). - Bug fix: Option -nmfe is set automatically if either -run_on_sge_nmfe, -run_on_slurm or -run_on_lsf_nmfe is set. - Bug fix: Handle model files with two $PROBLEM with two different data files. PsN-3.5.3 would set same data file for both $PROB. - Bug fix: A more informative error message when raw results have comma as the delimiter, which is not supported by PsN. ___General___ - Option -wrap_data has been removed. Use $SIZES in NONMEM instead. - When tweaking initial estimates for a retry, PsN will use a different method to ensure that OMEGA and SIGMA are stricly diagonally dominant. In 3.6.2 PsN will, if necessary, decrease the magnitude of the off-diagonal elements. PsN 3.5.3 would increase the diagonal elements. ___execute___ - Removed obsolete option -prepend_model_to_lst - New option: By default execute copies the data file to the NM run directory. By setting option -no-copy_data PsN will not copy the data file but instead use the full path to the data file in the model file. However, if the path is very long this can lead to a line-wrap in the $DATA record which will make NONMEM fail. ___gls___ - New option -set_simest to have different $EST for all steps, and different $SIM in input and simulation. Output computed shrinkage for all steps. Set NEW in default $SIM. - Bug fix: Remove $EST if $SIM ONLYSIM. - Bug fix: Handling of models with $ERROR. Fix errors in results computation. ___mimp___ - Bug fix: Ignore that data files set in $DATA of alternative models etc are missing at the start of the run. ___sse___ - Warn if .EQ. or .NE. is found in IGNORE/ACCEPT statements of alternative models. If found then warn that to NONMEM 1 is not equal to 1.000 (.EQ. and .NE. does string matching instead of numerical comparison) - sse will now accept to not estimate any models, only simulate and stop when -no-estim is set and no alternative models are given. - Bug fix: Ensure that random seed cannot be set to 0. ___scm___ - Linearized scm: If more than 50 items needed in $INPUT then set RFORMAT in $TABLE when generating derivatives data, to reduce risk of longer lines than 999 which NM7.2 cannot handle in a data file. - Bug fix: Handling of missing data in covariate data columns. - Bug fix: Linearized scm: If the number of items needed in $INPUT exceeds 50 then PsN will automatically set $SIZES instead of trying to automatically wrap data. - Bug fix: scm will now check that no parameter or covariate in the list of parameter-covariate interactions is named PAR or COV. PAR and COV are reserved words in scm and must only be used as place-holders in the [code] section. - Bug fix: Time-varying covariates (option -time_varying) now works. ___vpc/npc___ - By default vpc copies the data file to the NM run directory. By setting option -no-copy_data PsN will not copy the data file but instead use the full path to the data file in the model file. However, if the path is very long this can lead to a line-wrap in the $DATA record which will make NONMEM fail. - Bug fix: Cleaning without using globbing for file names. ___bootstrap___ - Code for automatic generation of R scripts for bootstrap results is no longer maintained and may not work. Functionality for handling bootstrap results is to be incorporated into Xpose. - Bug fix: Print informative error message if no runs pass the criteria for inclusion in the results computaions. ___RunRecord___ Bug fix: Handle tags that contain $. Before runrecord would break at $ in tags. ___psn_clean___ Remove temp_dir sub-directory of NM_run directories. ___update_inits___ New option -diagonal_dominance (can be abbreviated to diag). By default not set. If set then PsN will ensure that OMEGA and SIGMA are stricly diagonally dominant after updating by decreasing the magnitude of the off-diagonal elements if necessary. This option is useful in the cases when NONMEM prints OMEGA or SIGMA to the lst-file which NONMEM itself will not accept as input OMEGA or SIGMA because the matrix is not positive definite. ___extended_grid___ New userguide.