Homec4science
Diffusion Lenstool-HPC (lenstool_mpi) Browse

R1448/ChiBenchmark/lenstoolchi/perl/bayesResults.pl
c99f2e455a51lenstool_mpi

Last Change
Raw File

bayesResults.pl
View Options

#! /usr/bin/perl
# Read the bayes.dat file and print on STDOUT the parameter mean values and
# standard deviation.
#
# Syntax : bayesResults.pl [bayes.dat] [scale=<arcsec_to_kpc>]
#
use Term::ANSIColor;
use IO::Handle;
# The 3 sigma errors corresponds to 3 times the std deviation.
$PI = 3.1415927;
do "$ENV{'LENSTOOL_DIR'}/perl/median.pl";
do "$ENV{'LENSTOOL_DIR'}/perl/histogram.pl";
do "$ENV{'LENSTOOL_DIR'}/perl/rescaleBayes.pl";
$file = "bayes.dat";
$scale = 1;
while( @ARGV > 0 )
{
if( $ARGV[0] =~ /scale=/ )
{
@fld = split /=/,$ARGV[0];
$scale = $fld[1];
}
else
{
$file = $ARGV[0];
}
shift @ARGV;
}
if( $scale != 1. )
{
rescaleBayes($file, $scale);
$file = "bayes_rescaled.dat";
}
open(bayes, $file) || die "ERROR : $file not found.\n";
$nVal=0;
$bestline=0;
$bestchi2=1e100;
$lhood=0; # chi2 mode by default
while($line = <bayes>)
{
$line =~ s/\r$//; # change DOS to UNIX fileformat
chop($line);
if( $line =~ /#/ )
{
push @param, $line;
$lhood=1 if( $line =~ /Lhood/ );
} else
{
@fld = split / /, $line;
# Find the best chi2
$tmp = (($lhood==1)?-1:1)*$fld[1];
if( $bestchi2 > $tmp )
{
$bestchi2 = $tmp;
$bestline = $nVal;
}
for( $i = 0; $i <= $#fld ; $i++ )
{
$$values[$i][$nVal] = $fld[$i];
}
$nVal++;
}
}
close(bayes);
$nParam = $#param + 1;
printf "Read $nParam columns and $nVal lines\n";
printf "Param (nbin): <median> <best> <mode> <gausserr> <asymerr> (68%)\n";
for( $i = 1; $i < $nParam; $i++ )
{
# Set the best value
$best = $$values[$i][$bestline];
# Compute the mean value
$mean = 0;
for( $j = 0 ; $j < $nVal ; $j++ )
{
$mean += $$values[$i][$j];
}
$mean /= $nVal;
# Compute the median value
for( $j = 0 ; $j < $nVal ; $j++ )
{
$list[$j] = $$values[$i][$j];
}
#$median = median($nVal, \@list );
@slist = sort{ $a <=> $b } @list;
$median = $slist[$nVal/2];
# Compute asymmetric error bars
# Compute the Freedman-Diaconis bin size
$binsize = $slist[$nVal*0.75] - $slist[$nVal*0.25];
$binsize *= 2.*$nVal**-0.3333;
$binsize = 1 if( $binsize == 0);
$nbin = ($slist[$#slist]-$slist[0])/$binsize;
$nbin = 1 if( $nbin == 0);
($x, $histo) = histo($nVal, $nbin, \@list);
@shisto = sort{ $a <=> $b } @{$histo}; $hmax = $shisto[$#shisto];
# find the index of the largest bin
$hmaxid = 0;
$hmaxid++ while( $hmax != $histo->[$hmaxid] );
$mode = $x->[$hmaxid];
# find the mode index in <@list>
@slist = sort{ $a <=> $b } @list;
$modeid = 0;
$modeid++ while( $slist[$modeid] < $mode && $modeid < $nVal );
$modeid--;
# find 68% to the left of <modeid>
$eminid = $modeid - $modeid*0.68;
#$eminid-- while( $modeid - $eminid < $nVal*0.68 && $eminid > 0 );
$emin = $mode - $slist[$eminid+1];
# find 68% to the right of <modeid>
$emaxid = $modeid + ($nVal - $modeid)*0.68;
#$emaxid++ while( $emaxid - $modeid < $nVal*0.68 && $emaxid < $nVal );
$emax = $slist[$emaxid-1] - $mode;
# Compute the stddev value (bias corrected variance)
$stddev = 0;
for( $j = 0 ; $j < $nVal; $j++ )
{
$stddev += ($$values[$i][$j] - $mean)*($$values[$i][$j] - $mean);
}
$stddev /= $nVal - 1;
$stddev = sqrt( $stddev );
# Do not print Evidence line
#next if( $param[$i] =~ "Evidence" );
print color 'red' if( $hmaxid==0 || $hmaxid==int($nbin-1) );
if( $median < 1e4 )
{
$line = "%s (%.0f): %.4f %.4f %.4f +-%.4f +%.4f -%.4f (68\%)";
}
else
{
$line = "%s (%.0f): %.4e %.4e %.4e +-%.4e +%.4e -%.4e (68\%)";
}
printf $line, $param[$i], $nbin, $median, $best, $mode, $stddev, $emax, $emin;
print color 'reset';
print "\n";
}
c4science · Help