R help - Mar 2005 - Rounding parameter values in genoud(), Rgenoud package

I would like to limit the significant figures of the calibrated
parameters determined by genoud() in the Rgenoud package.  Below is some
example output, where column 1 is model run number, columns 2-7 are the
parameter values, and columns 8-12 are model fit statistics.  I would
like genoud to internally limit parameters to 4 decimal places as shown
in this output.  It is clear that the function is generating many
parameter sets that are identical after rounding.  Can I impose rounding
on the function and thereby lessen processing time?  The only function
argument that seems related is the solution.tolerance, but this is not
for the parameter values.

Run  par1    par2     par3     par4    par5     par6     Bias    MAE
R2       E2       E1'
507 0.0239 0.0219 0.0267 0.0274 0.0283 0.0245 -0.0112 0.0804 0.9994
0.9994 0.9792
508 0.0239 0.0219 0.0267 0.0274 0.0283 0.0245 -0.0112 0.0804 0.9994
0.9994 0.9792
509 0.0239 0.0219 0.0267 0.0274 0.0283 0.0245 -0.0112 0.0804 0.9994
0.9994 0.9792
510 0.0239 0.0219 0.0267 0.0274 0.0283 0.0245 -0.0112 0.0804 0.9994
0.9994 0.9792
511 0.0239 0.0219 0.0267 0.0274 0.0283 0.0245 -0.0112 0.0804 0.9994
0.9994 0.9792
512 0.0239 0.0219 0.0267 0.0274 0.0283 0.0245 -0.0112 0.0804 0.9994
0.9994 0.9792
513 0.0239 0.0219 0.0267 0.0274 0.0283 0.0245 -0.0112 0.0804 0.9994
0.9994 0.9792
514 0.0239 0.0219 0.0267 0.0274 0.0283 0.0245 -0.0112 0.0804 0.9994
0.9994 0.9792
515 0.0239 0.0219 0.0267 0.0274 0.0283 0.0245 -0.0112 0.0804 0.9994
0.9994 0.9792
516 0.0239 0.0219 0.0267 0.0274 0.0283 0.0245 -0.0112 0.0804 0.9994
0.9994 0.9792

Thanks,
Scott Waichler
Senior Research Scientist
Pacific Northwest National Laboratory
Richland, WA   USA
scott.waichler at pnl.gov

Hi Scott,
> I would like genoud to internally limit parameters to 4 decimal
> places as shown in this output.
Thanks for the question.  I don't know what your application is but
you may want to use the integer datatype (data.type.int=TRUE) and then
rescale the parameters in your function to provide the decimal
resolution you would like.  For example, set the genoud bounds so a
given parameter can range from -1,000,000 to 1,000,000 but in your fit
function divide the parameter by 10,000 so it actually ranges from
-100 to 100 and allows for four decimal places.  Judging from the
numbers you have printed, you could probably get away with the genoud
range being just -100,000 to 100,000 etc.

For example, set the "default.domains=100,000" or use the Domains
option to set individual parameter specific bounds; set
"data.type.int=TRUE"; and in your fit function:

FitFunction <- function(PARMS)
  {
    PARMS <- PARMS/10000
    [....your code....]
  }

Cheers, Jas.

=====================================Jasjeet S. Sekhon
Associate Professor
Harvard University
Institute for Quantitative
  Social Science
jasjeet_sekhon at harvard.edu
http://jsekhon.fas.harvard.edu/
Office: 617.496.2426 Fax: 617.507.5524
=====================================



> From: "Waichler, Scott R" <Scott.Waichler at pnl.gov>
> Date: March 2, 2005 1:01:01 PM CST
> To: r-help at stat.math.ethz.ch
> Subject: [R] Rounding parameter values in genoud(), Rgenoud package
>
>
> I would like to limit the significant figures of the calibrated
> parameters determined by genoud() in the Rgenoud package.  Below is
> some
> example output, where column 1 is model run number, columns 2-7 are the
> parameter values, and columns 8-12 are model fit statistics.  I would
> like genoud to internally limit parameters to 4 decimal places as shown
> in this output.  It is clear that the function is generating many
> parameter sets that are identical after rounding.  Can I impose
> rounding
> on the function and thereby lessen processing time?  The only function
> argument that seems related is the solution.tolerance, but this is not
> for the parameter values.
>
> Run  par1    par2     par3     par4    par5     par6     Bias    MAE
> R2       E2       E1'
> 507 0.0239 0.0219 0.0267 0.0274 0.0283 0.0245 -0.0112 0.0804 0.9994
> 0.9994 0.9792
> 508 0.0239 0.0219 0.0267 0.0274 0.0283 0.0245 -0.0112 0.0804 0.9994
> 0.9994 0.9792
> 509 0.0239 0.0219 0.0267 0.0274 0.0283 0.0245 -0.0112 0.0804 0.9994
> 0.9994 0.9792
> 510 0.0239 0.0219 0.0267 0.0274 0.0283 0.0245 -0.0112 0.0804 0.9994
> 0.9994 0.9792
> 511 0.0239 0.0219 0.0267 0.0274 0.0283 0.0245 -0.0112 0.0804 0.9994
> 0.9994 0.9792
> 512 0.0239 0.0219 0.0267 0.0274 0.0283 0.0245 -0.0112 0.0804 0.9994
> 0.9994 0.9792
> 513 0.0239 0.0219 0.0267 0.0274 0.0283 0.0245 -0.0112 0.0804 0.9994
> 0.9994 0.9792
> 514 0.0239 0.0219 0.0267 0.0274 0.0283 0.0245 -0.0112 0.0804 0.9994
> 0.9994 0.9792
> 515 0.0239 0.0219 0.0267 0.0274 0.0283 0.0245 -0.0112 0.0804 0.9994
> 0.9994 0.9792
> 516 0.0239 0.0219 0.0267 0.0274 0.0283 0.0245 -0.0112 0.0804 0.9994
> 0.9994 0.9792
>
> Thanks,
> Scott Waichler
> Senior Research Scientist
> Pacific Northwest National Laboratory
> Richland, WA   USA
> scott.waichler at pnl.gov
>
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide!
> http://www.R-project.org/posting-guide.html
>