BPZ Quick Start

• UDFtest.cat -- photometric catalog (including spec-z, if available)
  
• format properly:
• m =  27.3, dm = 0.02 -- example of a detection
• m =  99.0, dm = 28.4 -- example of a non-detection: dm = 1-sigma detection limit
• m = -99.0, dm = 0    -- non-detection
• UDFtest.columns -- linking the catalog columns to the:
  
• ~/bpz-1.99.3/FILTER/*.res -- filter response curves
• 2 columns: wavelength in Angstroms; response (normalization not necessary)

1. Run BPZ
• bpz UDFtest.cat -INTERP 2
• INTERP: interpolates 2 SEDs between each adjacent pair
2. Check for photometric offsets in individual filters!
• Printed as output
• Any offsets may be input in the .columns file (zp_offset)
  
• Re-run BPZ if necessary!
  
3. Produce a "final" output file
   
• python $BPZPATH/bpzfinalize.py UDFtest
• Includes calculation of modified chisq (chisq2) plus different formatting
  
4. Important: Check your results including SED fits and P(z)
   
• python $BPZPATH/plots/webpage.py UDFtest
• produces html/index.html by default
• python $BPZPATH/plots/webpage.py UDFtest i1-5
• webpage of first five objects only
• for other input options, see the comments in the header of ~/bpz-1.99.3/plots/webpage.py
  
• Or to look at individual plots:
  
• python $BPZPATH/plots/sedplot.py UDFtest <NNN> <iMMM>
  
• assumes you set -CHECK yes and produced UDFtest.flux_comparison
  
• will plot all objects in turn unless you pick the MMMth object or the object with id = NNN
  
• python $BPZPATH/plots/probplot.py UDFtest <NNN>
5. Prune your catalog, keeping only those with:
• chisq2 < 1: Good SED fits (worse fits are generally unreliable)
• Ideally you will consider the full P(z) for each object.  If not possible, you may wish to prune:
  
• ODDS >= 0.95, if desired: those objects for which zb is probably correct (no secondary solutions)
  

• Photometric recalibrations for each filter (if large, re-run BPZ)
• UDFtest.bpz,  UDFtest_bpz.cat
• zb, tb:  Most likely redshift and spectral type
  
• zbmin..zbmax:  95% confidence interval
• ODDS:  Likelihood that zb is correct (within 0.10).  Low odds indicate multiple possible solutions.
• chisq2:  "Modified" chisq.  Objects with chisq2 > 1 may have unreliable SED fits.
  
• zml, tml:  Maximum Likelihood results (with "flat" unphysical prior).  Generally worse.
  
• UDFtest.probs -- P(z) for each galaxy
  
• html/index.html -- webpage to check results (including SED fits and P(z))
• UDFtest.flux_comparison -- one row per galaxy with columns:
  
• ID
• mag 
• zb (photo-z)
• tb (type)
• am (factor used to multiply ft to match fo)
• ft[nf] (model fluxes * am; nf columns where nf = # of filters)
• fo[nf] (observed fluxes)
• efo[nf] (observed flux uncertainties)