MCCLASS / CLASSIFY

Purpose:

Classification of a CADIS-format object table: On the basis of colour related information, the probability of being a star, a white dwarf or blue HB star, a galaxy or a quasar is calculated for every object. Together with the morphology, and potentially the variability, a final interpretation is reached.


Syntax:
	MCCLASS/CLASSIFY [flux-table] [me] [par] [list] [min] [limit] [var]

	   flux-table = Name of the CADIS-format object table to be classified
			(default: flux_SURVEY.tbl)
	   me         = Minimum error assumed for colour indices
			(default: 0.05)
	   par	      = Name of a parameter file determining the colors 
			(default = feature_combo.par)
	   list       = Flag for MIDAS-listing of detailed solutions 
			(List or default = NoList) 
	   min        = Required minimum probability for classifiability in %
			(default = 75.0)
	   limit      = Maximum sigma of fit allowed for classification
			(default = 10.0) 
           var        = Variability prior to be used? 
                        (default = varyes)


Examples:	MCCLASS/CLASSIFY 
		MCCLASS/CLASSIFY flux16h_Dez.tbl
		MCCLASS/CLASSIFY flux9h_27Apr98.tbl ? fea_4filt.tbl
		MCCLASS/CLASSIFY flux9h_27Apr98.tbl 0.03 ? List
		MCCLASS/CLASSIFY flux16h_Dez.tbl ? ? ? 90.0 
Remarks:

The object table has to be run through MCCLASS/COLOURS first in order to calculate the colour indices and their errors for all objects. Furthermore, a :class-column is required as well as a number of other columns which are to incorporate the final interpretation (see below). The parameter file defining the comparative colours has to be in the directory where the command is called.

The probability values for the classification are based on a template-fit procedure making use of three different kinds of information which are:
  • the previous knowledge in the form of colour libraries of stars, white dwarfs, galaxies and quasars
  • the measured value for the colours of the objects
  • the measuring error for the colours of the objects
Moreover, the classification estimates the redshifts for the extragalactic classes which are determined with two procedures: (1) the minimum error variance method (MEV) and (2) the maximum likelihood method (ML). During the process, the programme first calculates the solution for the star interpretation, estimates subsequently the WD/BHB/sdB case, the galaxy case and then the quasar case. Then, the probabilities for the individual classes are compared and the final interpretation is entered into the result columns.
Explanation of the result columns:

1. General results
  • prob_star probability in %, that an object is a star
  • prob_wd probability in %, that an object is a wD/BHB/sdB star
  • prob_gal probability in %, that an object is a galaxy
  • prob_qso probability in %, that an object is a quasar The sum of the three probabilities is 100%. It is assumed that there exists no other kind of object.
  • MC_class Final classification of the object determined by using colour and morphology as well as by respecting the minimum probability required (stars and WD/BHBs combined need to reach p_min): Star = definite stars (colour of star, stellar shape if z<0.2) WDwarf = definite WD/BHB/sdB star (colour of WD/BHB/sdB, shape if z<0.2) Galaxy = definite galaxies (colour of galaxy, shape irrelevant) Galaxy (Star?) = most likely galaxy at z<0.2 (star colour, but extended shape and z<0.2 - overlap in colour space!) Galaxy (Uncl!) = colour undecided (statistically almost always a galaxy) QSO = definite QSOs (colour of QSO, stellar shape) QSO (Gal?) = colour of QSOs, extended shape (low-L AGN or contaminating galaxy) Strange Object = very strange spectrum (strong photometric artifacts or uncorrected strong variability or unusual spectrum, e.g. strong ELs)
2. Minimum error variance results for the classified interpretation:
  • MC_z redshift of the object corresponding to classification. z-value of MEV estimate, if galaxy or quasar, *=null, if star or WD or no MEV estimate possible.
  • MC_sz error estimate for redshift (sigma of p(z) distribution)
  • MC_z_bin interval, from which the solution for :MC_z is taken: 0: There is no solution. p(z) is nearly evenly distributed. 10: There is exactly one solution. 21: There are two solutions. The one at lower z is given. It is more likely. 22: There are two solutions. The one at higher z is given. It is more likely.
  • MC_gSED spectral type of the object, if galaxy (0..5): SED = 0.0 .. 1.2 means dust-free (mostly bulges) SED = 1.2 .. 5.0 means increasing extinction
  • MC_sgSED error estimate for the SED (sigma of distribution)
  • MC_gSED2 spectral type of the object, if galaxy (0..59): from 0 to 59 the mean age of the population increases from 50 Myr to 15 Gyr for a tau=1 Gyr exponential SFH SED = 0..25 means starbursts SED = 25..42 means typical disk galaxies SED = 42..59 means typical bulges
  • MC_sgSED2 error estimate for the SED2 (sigma of distribution)
3. Maximum likelihood results for the classified interpretation:
  • MC_z_ml redshift of the object corresponding to classification. z-value, if galaxy or quasar, *=null, if star.
  • MC_gSED_ml spectral type, if galaxy (0..5), else *=null
  • MC_gSED2_ml spectral type, if galaxy (0..59), else *=null
4. Maximum likelihood results for the nearest library object:
  • MC_nextlib number of the library object. The name of the library is given in the descriptor MC_library{number}/c/1/20 (descriptors are MC_library1, MC_library2 and MC_library3).
  • MC_next_0 parameter index 0 of the object (defines position of
  • MC_next_1 ... 1 the object within
  • MC_next_2 ... 2 the library array)
  • MC_next_3 ... 3
  • MC_dist distance to the nearest library object (kind of sigma), needs rescaling. Presently, strange objects are best selected by MC_dist.gt.6.0
Further results are used for internal success checks and detailed analyses:
  • sdmg distance from the star library (units as in MC_dist)
  • wdmg distance from the WD/BHB library (units as in MC_dist)
  • gdmg distance from the galaxy library (units as in MC_dist)
  • qdmg distance from the quasar library (units as in MC_dist)
  • spI probability integral of the star library
  • wpI probability integral of the WD/BHB library
  • gpI probability integral of the galaxy library
  • gpIa partial integral of the lower z-interval
  • gpIb partial integral of the upper z-interval
  • qpI probability integral of the quasar library
  • qpIa partial integral of the lower z-interval
  • qpIb partial integral of the upper z-interval
  • stype MEV estimate of the star type (meaningless since 1999!)
  • sstype error of this estimate (meaningless since 1999!)
  • sdmg3 parameter 3 of the nearest star (= row number in the star library)
  • wdmg3 parameter 3 of the nearest WD/BHB (= row number in the WD/BHB library)
  • gz MEV estimate of the galaxy redshift
  • gsz error of this estimate
  • gza estimate of the redshift in the lower interval
  • gsza error of this estimate
  • gzb estimate of the redshift in the upper interval
  • gszb error of this estimate
  • gSED MEV estimate of the galaxy SED
  • gsSED error of this estimate
  • gSEDa estimate of the SED in the lower interval
  • gsSEDa error of this estimate
  • gSEDb estimate of the SED in the upper interval
  • gsSEDb error of this estimate
  • gSED2 MEV estimate of the galaxy SED2
  • gsSED2 error of this estimate
  • gSED2a estimate of the SED2 in the lower interval
  • gsSED2a error of this estimate
  • gSED2b estimate of the SED2 in the upper interval
  • gsSED2b error of this estimate
  • gdmg0 parameter 0 of the nearest galaxy
  • gdmg1 ... 1
  • gdmg2 ... 2
  • gdmnr row number of this galaxy in the galaxy library
  • qz MEV estimate of the quasar redshift
  • qsz error of this estimate
  • qza estimate of the redshift in the lower interval
  • qsza error of this estimate
  • qzb estimate of the redshift in the upper interval
  • qszb error of this estimate
  • qalph MEV estimate of the quasar spectral index
  • qsalph error of this estimate
  • qepsi MEV estimate of the quasar emission line strength
  • qsepsi error of this estimate
  • qdmg0 parameter 0 of the nearest quasar
  • qdmg1 ... 1
  • qdmg2 ... 2
  • qdmnr row number of this quasar in the quasar library
ATTENTION: All integer numbers follow C-convention, first entry = 0


Colour libraries: (as of 23 June 2003)

For stars, WD/BHBs/sdBs, galaxies and quasars, there is one colour library, respectively, containing about 70 colour indices including those for the SDSS, ODT, CADIS and COMBO-17 surveys, as well as Johnson, WFPC2 and other bands. For the classification only those colour indices are used which can be found simultaneously with identical names in three places: the colour library, the object table and the parameter file. In general, the colour libraries are organized as four-dimensional arrays; in the case of a simpler data structure, some dimensions collapse. In analogy to the sequential storage of array variables in programming languages, the library arrays are sequentially organized as MIDAS tables. The index of the parameter 0 has the slowest upward count, the index for parameter 3 changes in every line. Parameter 0 is generally reserved for the redshift being the only one which, in the case of fuzzy distributions, can split into two partial intervals by the classification programme. The three colour libraries currently recommended are found in /photo/user/tbl/COLORLIB/ and are explained here:
  • fG_UK96.tbl

    The star library, derived from the UK spectra of the Pickles catalogue (1998 PASP 110, 863). The 96 stars of spectral types F,G,K and M having luminosity classes I to V are used. The library has an unsorted structure.

    Weakness: It contains no L-stars and no white dwarf/M-dwarf-binaries.

  • fG_WDwarf.tbl

    The WD/BHB/sdB library, derived from the theoretical spectra and provided by Detlev Koester (more details follow soon) covering DA types with temperatures ranging from 6000 K to 40000 K and surface gravities from log g = 9 to log g = 6 (sub-dwarfs sdB/sdA). The library has an unsorted structure.

  • fG_KICA100_M2002.tbl

    The older alternative:
    The galaxy library, derived from the template spectra by Kinney, Calzetti et al. (1996, ApJ 467, 38). The ten templates for the different galaxy-SEDs (E up to S1 = *B1) from 125 nm up to 800 nm have been adopted almost unchanged (assumed artefacts were removed at 290 to 340 nm, at 540 nm and beyond 680 nm). Towards the K-band the F_nue spectra were extrapolated by a simple power law, with the spectral index -0.333 for the types E to Sb, and with the spectral index -1.0 for star burst galaxies. Within the flux space, the ten templates were extended to 100 spectral types  by interpolation. The library is organized as a two-dimensional array. Parameter 0 is the redshift, which assumes 460 values in equidistant steps of 0.002 on a log(1+z)-axis within the interval from z = 0.00 up to 1.502. Parameter 1 is the spectral type, whose translation key into local morphological types is explained above. Thus, the library contains 46000 objects.

    Weakness: The one-dimensional spectral type cannot perfectly reproduce the different ratios of lines which, in the case of galaxies with two or more emission lines in medium-band filters, may lead to a wrong solution (e.g. z = 0.25 vs. 0.66 in CADIS filter set; here, the emission line analysis should offer a more reliable result). Also, Seyfert-2-galaxies have occasionally different combinations of continuum slopes not realized in this library, so their redshifts can be wrong.

  • fG_REx1000_SMC6.tbl

    The newest galaxy library:
    A galaxy library calculated with the PEGASE population synthesis package in a 2-D format with 60 ages times 6 extinxtion levels. The ages range from 50 Myr to 15 Gyr and contain some interpolated steps besides the age steps produced by PEGASE itself. The SFR follows an exponential decline with a tau of 1 Gyr. The model involves standard parameters, a Kroupa IMF, an initial metallicity of 0.01 and is extinction-free. The extinction is applied afterwards as a screen following the SMC law defined by Pei (1992) with six values of E(B-V)=[0.0,0.1...0.5]. The SMC law is most appropriate for higher-z galaxies, while at z<0.5 the 220 nm bump of higher metallicity galaxies is not visible in our filter set anyway, so its absence from the SMC law is irrelevant. 177 steps in redshift cover the range from 0.0..1.4 at 0.005 resolution on a log(1+z) scale.

  • fG_QSO_sdss05.tbl

    The quasar library, derived from the SDSS template spectrum. The template is varied in intensity, added to a power law continuum and multiplied by a redshift-dependent throughput function modelling the Hydrogen absorption bluewards of the Lyman alpha line. The library is organized as a three-dimensional array. Parameter 0 is the redshift which adopts 155 values in equidistant steps of 0.01 on a log(1+z)-scale within the interval z = 0.504 up to 5.96. The template does not cover sufficiently red wavelengths to be used at lower z. Parameter 1 ist the spectral index of the continuum and assumes 20 different values. Parameter 2 is the intensity of the emission line shape and assumes eight values of intensity relative to the mean template (more details follow soon). Thus, the library contains 30600 objects.

    The differentiation between galaxy and QSO depends on the ratio of nuclear luminosity to host luminosity. Effectively, it appears, as if all objects of M_B < -21.5 are classified as QSO, provided their apparent magnitudes are bright enough, including a few lower-luminosity objects.

Parameter files:

The parameter files contain information about the colour indices, the libraries to be used for comparison and the magnitude used for a classification improved by a MAP approach taking class abundance known a-priori into account. This file allows to switch off specific colour indices for the classification or switch off the MAP1-classification stage or use custom-made color libraries. Example:
!	feature_cadis.par	C. Wolf 06.03.2002
!	example parameter file for CADIS colors
!
feature=:bmr
feature=:rmi
feature=:imk
!
feature=:b531mr
feature=:b612mr
!  feature=:rmb700
feature=:b752mi
feature=:imb855
feature=:imb909
feature=:bmv465
!
MAP_mag=:_M_I__E
!
library1=/photo/user/tbl/COLORLIB/fG_UK96.tbl
library2=/photo/user/tbl/COLORLIB/fG_KICA100_Feb02.tbl
library3=/photo/user/tbl/COLORLIB/fG_QSO_sdss.tbl
library4=/photo/user/tbl/COLORLIB/fG_WDwarf.tbl

Lines beginning with ! are ignored as they are regarded as comment lines. The command feature:= announces a comparative feature to the classification programme. Actually it is only used, when it is also available in both the library and the object table. The command library{i}= announces the path and filename of libraries to be used. Library1 must be for stars, library2 for galaxies, library3 for quasars and library4 for WD/BHBs. If the command for any library is omitted, the corresponding recommended library (see above) is used.

List option:

The list option is a remnant from testing times by means of which the classification of individual objects can be traced. The output of the programme is then also recorded in the MIDAS logfile. If this option is used, the programme is processed more slowly.



Back to COMBO-17 index page


Last update Oct 30, 2003, CW