Dear Carlos, This warning is triggered when relative standard errors on a parameter exceed a user-supplied limit. These limits can be specified in the "General" tab of the Options dialog, for THETA, OMEGA and SIGMA parameters (which have defaults of 0.3, 0.5 and 0.5, respectively).
This actually qualifies as a bug report, since I notice that the limit options are incorrectly labelled as "% Coefficient of variation". "Solving" them might be non-straightforward, and is fundamentally a model-building issue - large standard errors are associated with parameters that are poorly estimated, and may indicate potential overparameterization or model misspecification. Without knowing details of the model in question, it's difficult to be much more specific, but in general it's helpful to try dropping poorly estimated parameters from the model one at a time to assess the impact of doing so on the the other parameter estimates, the objective function value and the diagnostic plots. If such effects are negligible, the poorly-estimated parameter is probably not needed. Best Justin 2009/6/9 Carlos Hoyo <citocr...@cinvestav.mx> > Dear Justin, > How to solve the "Large SEs" reported by Census? > > Saludos, > > -- > > Dr. Carlos Hoyo Vadillo > Sec. Ext. Farmacología > Cinvestav > Av. IPN 2508 > San Pedro Zacatenco > 07360 México DF > > Tel: +52-55-5061-3800 ext 5438 > Cel. 55-385 22 888 > > citocr...@cinvestav.mx > > > > > > > > > On Jun 2, 2009, at 5:09 AM, Justin Wilkins wrote: > > Dear Carlos, > > In principle, porting to other platforms is a wonderful idea, but currently > time constraints and the dependence of Delphi on Windows make this > impossible for a one-man spare-time development team like myself in the > short term. Of course, the source code is freely available - and if someone > would like to give this a go in something like Lazarus (the cross-platform > Free Pascal environment) or, even better, create a C++ implementation using > something like Qt, that would be wonderful... > > Best regards > Justin > > 2009/5/26 Carlos Hoyo <citocr...@cinvestav.mx> > >> Dear Justin, >> Plans for a Mac Os or linux version? >> >> Saludos, >> >> -- >> >> Dr. Carlos Hoyo Vadillo >> Sec. Ext. Farmacología >> Cinvestav >> Av. IPN 2508 >> San Pedro Zacatenco >> 07360 México DF >> >> Tel: +52-55-5061-3800 ext 5438 >> Cel. 55-385 22 888 >> >> citocr...@cinvestav.mx >> >> >> >> >> >> >> >> >> On May 17, 2009, at 6:10 PM, Justin Wilkins wrote: >> >> Dear all, >> >> Census 1.1 has been released to beta. This version includes a number of >> enhancements: >> >> * The R(D)COM library has been removed, fixing some nasty >> incompatibilities with R version 2.8.0 and above (affecting Xpose plotting). >> It has been replaced with R-based file output, including support for PDF, >> JPG, PNG, BMP, Postscript and Xfig formats. Files have the additional >> advantage of being retained after the plot window is closed. >> * An R interaction layer and console window, controlled by a taskbar >> button, has been added. >> * MD5 file fingerprints are now optionally collected for NONMEM input and >> output, and checked for changes whenever accessed within Census. >> >> You can get it from http://census.sourceforge.net/, as usual. Comments, >> suggestions, support requests, bug reports and the like are as always very >> welcome, and can be sent to the Census mailing list - registration is >> provided at http://census.sourceforge.net/docs.htm. >> >> Census is a comprehensive project manager for NONMEM, providing detailed >> summary, comparison and overview of the runs comprising a given project, >> including the display of output data, post-run processing, and rapid >> diagnostic plots through tight integration with Xpose. While it runs only >> under Windows, it can read any NONMEM output to which Windows has access to >> via drive mapping. >> >> Census is, and will remain, free software under the Mozilla Public >> License, version 1.1. >> >> Last but not least, while this is a stable release, it's still technically >> a beta, incorporating a brand new backend, and it has not been as rigorously >> tested as a full release might be. Please bear this in mind before deleting >> your old versions of the software... >> >> Best >> Justin >> >> -- >> Justin Wilkins >> ----------------------------- >> Colmarerstrasse 31 >> 4055 Basel >> Switzerland >> ----------------------------- >> Email: justin...@gmail.com >> Skype: kestrel_za2002 >> Mobile: +41 76 561 0949 >> >> >> > > > -- > Justin Wilkins > ----------------------------- > Colmarerstrasse 31 > 4055 Basel > Switzerland > ----------------------------- > Email: justin...@gmail.com > Skype: kestrel_za2002 > Mobile: +41 76 561 0949 > > > -- Justin Wilkins ----------------------------- Colmarerstrasse 31 4055 Basel Switzerland ----------------------------- Email: justin...@gmail.com Skype: kestrel_za2002 Mobile: +41 76 561 0949
