Locked History Actions

DeepZot/Baofit

BAOFIT

Code is hosted at github.

History

Click on a date for details:

  • 24-Mar-17:

    • Use input parameter to set the matter density value in the velocity shift for cross-correlation fits
  • 6-Mar-17:

    • Implement r-space quasar radiation effects model in k-space correlation model
  • 2-Feb-17:

    • Rework metal model implementation and add support for fitting CIV correlations
  • 12-Dec-16:

    • Modify and add options for k-space smoothing models
  • 26-Nov-16:

    • Fix bug in repeating the k-space transformation in the cross-correlation fit
  • 4-Jun-16:

    • Add support for HCD model in the cross-correlation fit
  • 30-May-16:

    • Modify HCD model by adding one more fit parameter
    • Modify UV fluctuation model by redefining the UV bias parameter
  • 25-May-16:

    • Add support for k-space HCD model
  • 19-May-16:

    • Add support for alternative pixelization smoothing
  • 15-Mar-16:

    • Add missing return statements in correlation models
  • 8-Mar-16:

    • Add separate options for using the combined fit parameters beta*bias and aperp/apar
    • Fix a bug in redshifting of the additive broadband for the cross-correlation fit
  • 6-Mar-16:

    • Add redshift criterion, set by dzmin parameter, for repeating the transformation to r-space.
    • Add additional RSD criteria for repeating the transformation to r-space. Fixes bug that ignored changes in beta2*bias2.
    • Update zeff parameter handling to take the value from the data, if not set separately.
  • 26-Feb-16:

    • Add option to apply a broadband function before applying the distortion matrix
  • 25-Feb-16:

    • Change the definition of a combined fit parameter to aperp/apar
  • 19-Feb-16:

    • Add support for fitting parameters of the non-linear correction
  • 14-Feb-16:

    • Add support for interpolation-based metal correlation model in the cross-correlation fit
  • 11-Feb-16:

    • Add support for distortion matrix in the cross-correlation fit
  • 10-Feb-16:

    • Add support for k-space UV fluctuation model
    • Add zeff redshift parameter to be used by the non-linear correction and UV fluctuation model
    • Update non-linear correction parameters to published values in AriƱo-i-Prats et al. 2015
  • 27-Jan-16:

    • Fix a bug in the call to the metal correlation model for the distortion matrix option
  • 25-Jan-16:

    • Add option to use an index-based metal correlation model
  • 22-Jan-16:

    • Set the distortion matrix name to be the same as the data name by default
  • 22-Jan-16:

    • Add option to fit for combined parameters (beta*bias and apar/aperp)
  • 16-Dec-15:

    • Add print to screen of the dilation limit at runtime error
  • 16-Dec-15:

    • Update sampling analysis to catch runtime error and print failed fits to output file. Fix for issue #19

    • Add print to screen of the dilation limit at runtime error. Fix for issue #18

  • 13-Dec-15:

    • Add support for distortion matrix in k-space multipole model
    • Add optional fit parameter describing the smoothing scale set by the pixel size
  • 26-Nov-15:

    • Add support for non-linear correction models in k-space multipole model
    • Needs updated version of cosmo
  • 25-Nov-15:

    • Add new k-space hybrid transformation model, including k-space distortion model, based on cosmo/DistortedPowerCorrelationHybrid
    • Needs updated version of cosmo
  • 16-Oct-15:

    • Add support for improved metal correlation model to fit for bias and beta of the relevant metal absorptions by Si II and Si III
    • Add support for using a custom binning grid of non-uniform sampling points read from file
    • Needs updated version of likely
  • 1-May-15:

    • Add support for r-space model of metal correlations
  • 23-Apr-15:

    • Add new k-space FFT models, including k-space distortion model, based on cosmo/DistortedPowerCorrelationFft
    • Add support for non-linear correction models
    • Needs updated version of cosmo
    • Add BOSSDR11LyaF_fft config file to reproduce Blomqvist et al 2015 fit
    • Add BOSSDR11LyaF_fft scan file and update existing DR11 scan files to new scan grid
    • Update existing DR11 config files to new scan grid
    • Update instructions in data/README.txt file
  • 13-Mar-15:

    • Update baofit to dump minimum chisquare to output file
  • 30-Jan-15:

    • Update bootstrap to print failed fits to output file. Fix for issue #15

  • 19-Dec-14:

    • Add BOSSDR11LyaF_k data, cov, scan files
    • Add BOSSDR11LyaF_k config file to reproduce Delubac et al 2015 fit
    • Update instructions in data/README.txt file
    • Fix minor bug with reference redshift in BOSSDR11QSOLyaF_k config file
    • Extend linear matter power spectrum templates DR9LyaMocks and DR9LyaMocksSB to include larger k values
  • 16-Sep-14:

    • Updates to support the new clang compiler included with Mac OS 10.9 (Mavericks).
    • You will also need the latest likely and cosmo packages for Mavericks.
    • Backwards compatible with previous versions of Mac OS X.
  • 26-Feb-14:

    • Fixes for issues #11 and #12 reported by Nicolas

  • 5-Feb-14:

    • Add support for k-space models based on cosmo/DistortedPowerCorrelation
    • Add config files that reproduce published r-space fits with k-space models (and allow floating non-linear effects)
    • Remove CAMB output files in models/ that were accidentally added earlier
    • Needs updated versions of likely and cosmo
  • 10-Jan-14:

    • Update mean redshift of Busca 2013 data in INI file
    • Add support for fractional powers in r for BroadbandModel

    • Implement Busca 2013 eqn(24) as default fit config
    • Use finer scan grid sampling for Busca 2013 fit and update .scan file in data/
  • 8-Jan-14:

    • Add new model template files with extended r coverage (DR9Mocks + PlanckWP)
    • Add data/ directory containing .data and .cov files for published results
    • Add data/README.txt with instructions for reproducing public results
    • Add text files with (aperp,apar,chisq) scan results for published results
    • Update Slosar 2013 data,cov files to match published baseline fit
    • Add Busca 2013 data,cov file provided by Nicolas
  • 18-Jul-13:

    • Add BOSSDR11QSOLyaF.ini config file, similar to existing BOSSDR9LyaF.ini
    • Standardize on LyaF (instead of LyAF) in filenames
    • Add cov-sample-size command line option to automatically rescale chisq icov by (N-n-2)/(N-1)
  • 14-Jun-13:

    • Re-enable support for fitting multipole datasets (more details here)

    • Multipole fits now write results in a format suitable for input to a multipole fit
    • General cleanup of obsolete parts of the code and the BOSS DR9 Lya legacy fit
    • Update demo/ with a companion wiki page

    • Minor bug fix that affected error on beta (except for cross correlation fits)
  • 13-May-13:

    • Update to use new likely features (needs latest version of likely)
    • Add --parameter-scan option that uses the new fit parameter binning
    • Changes to binning syntax are not backwards compatible (see the AbsBinning header comments for details)

    • Update binning specs in config files
  • 8-May-13:

    • Add optional rP,rT terms to the polynomial parameterization of distortion models. See the BroadbandModel header comments for details on how the command-line syntax for distortion models has been generalized to accommodate the new terms. Note that the original syntax is still supported, for backwards compatibility, but should now be prefixed with "r,mu,z=" with consistency with the other formats.

  • 15-Apr-13:

    • Add --load-wdata option equivalent to the --weighted option for cosmolib.
    • For backwards compatibility, the cosmolib --saved-format option should be accompanied by --load-icov since the previous hard-coded default has been removed.
  • 19-Mar-13:

    • Generalize linear-bias model in AbsCorrelationModel to work for cross correlations

    • Add fit parameter for radial velocity shift to allow for systematic errors in redshift
    • Include Andreu's quasar radiation model in BaoCorrelationModel

  • 21-Feb-13:

    • Generalize ComovingCorrelationData to be either (r,mu,z) or (rpar,rperp,z)

    • Change default mu-min from 0 to -1, to support cross correlation data with signed rpar and mu
    • Use new likely::createBinning to support cmd-line specification of arbitrary binning
    • Add cmd-line options comoving-polar, comoving-cartesian, axis1-bins, axis2-bins, axis3-bins, load-icov
    • Unify code to read Aln2 and comoving input data
  • 18-Dec-12:

    • Aln2 INI file updates:
      • trim unused dll bins (maxll changed from 0.27 to 0.10)
      • use final bband distortion models
      • synchronize baseline param config with Anze (including use of --decoupled)
      • update xi,pk configs to match fitting paper
    • Code updates:
      • migrate saveData, saveInverseCovariance from baofit::AbsCorrelationData to likely::BinnedData

      • use new likely methods to dump best-fit params and their covariance
  • 27-Nov-12:

    • Implement new BroadbandModel that can be used standalone or for multiplicative, additive distortion.

    • Distortions are configured with new cmd-line params dist-add, dist-mul, dist-r0.
    • Eliminate !BBand1 and the corresponding model files.
    • Eliminate !BBand2 (equivalent to dist-add=-2:0,0:4:2,0 in updated code).
    • Temporarily disable multipole-space support in BaoCorrelationModel.

    • Add cmd-line option --decoupled that only applies scales to BAO peak (by default, scales are now applied to everything, unlike in previous versions).
  • 25-Nov-12:

    • Add two options for including anisotropic non-linear effects (see models/README and config/Aln2.ini for details)
    • Add missing newlines at end of some model data files (has negligible effect on numerical results).
  • 17-Nov-12:

    • Add cmd-line options --compare-each, --compare-each-final, --fix-mode-scales, --project-modes-keep
    • dumpResiduals uses an externally provided combined dataset.
    • Improve reporting of exceptions.
    • Rename --llmin to --ll-min (!) and add ll-max, sep-min, sep-max cuts.

  • 24-Oct-12:

    • New models with isotropic non-linear broadening applied via a factor exp(-(S*k)^2/2) that multiplies P(k).
    • NL3,NL6,NL12 models are calculated with S = 3,6,12 Mpc/h.
    • Models exported by Mathematica do not have a final newline, so the last data point is not currently being used for interpolation (with negligible impact on fit results).
    • This should either be fixed in Mathematica or likely::readVectors(), but I am not making any change to the affected files here, for backwards compatibility.
  • 27-Sep-12:

    • Fail gracefully if we don't have enough samples for a bootstrap covariance estimate.
    • Always save work in progress on bootstrap covariance to allow offline combinations (see comments for CovarianceAccumulator::dump() for details on the format).

    • Always dump best-fit parameters in model-config format to <output-prefix>fit.config

    • Remove weighted=yes as default in Aln2 INI file since reading unweighted data is now generally more efficient (especially with scalar-weights enabled).
  • 25-Sep-12:

    • Add save-data option to save unweighed combined data vector.
    • Add bootstrap estimate of combined covariance (bootstrap-cov-trials).
    • Add scalar-weights option to assume covariances have same structure (correct but slow for now)
    • Add no-initial-fit option to skip initial fit and use input model-config instead (supersedes mcmc-reset)
    • Add compare-each option to print chi2 of each plate relative to their weighted avg.
    • Rename "mc" to "toymc" in options and code.
    • Better implementation of reuse-cov option for Aln2 format data.
    • Update to new likely BinnedDataResampler, BinnedData, CovarianceMatrix, CovarianceAccumulator.

  • 11-Sep-12:

    • Add a new mcmc-reset option that ignores the MINUIT covariance matrix and instead uses the model-config for MCMC.
    • Bug fix: llmin cut was not implemented correctly.
  • 10-Sep-12:

    • Updated likely pkg adds support for soft box priors.
    • Implement fix-aln-cov using hardcoded magic numbers (from Anze's pull/1 request)

    • Implement z-min, z-max cuts (from Anze's pull/5 request)

    • Refactor AbsCorrelationData and its subclasses to move all r,mu,z cuts into base class.

  • 8-Sep-12:

    • Add new P(k) interpolating mode PkCorrelationModel (see config/Aln2_pk.ini)

    • Implement second "demo" file format for Aln2 correlation estimates (see config/demo.ini)
    • Add options to generate and save a toy MC dataset in the new Aln2 format
    • Add a demo/ directory with sample toy MC Aln2 data representing 150x DR9 (see demo/README)
    • Refactor Xi,Pk correlation models to use linear bias stuff from base class (update BAO model next)
  • 29-Aug-12:

    • Fix BBand2 hexa for Xu files, since they have no hexa data
    • Update to new FitParameterStatistics API and add stats for MCMC

  • 29-Aug-12:

    • Remove obsolete flux_Aln2_ab.ini and regression.ini
    • Rename flux_X.ini to X.ini (since it works also with delta field inputs)
    • Standardize config sections in Aln2.ini, Aln2_xi.ini and Xu.ini
    • Update (rmin,rmax) defaults: (50,170) for BAO fits, (20,190) for spline fits.
    • Rename "BAO scale a,b" to "BAO alpha-parallel,perp" and change nominal values from 0 to 1.
    • Use ndump = 0 by default (gives much smaller analysis output files)
    • Rename "BAO scale" to "BAO alpha-iso", and "alpha-*" to "gamma-*" to match fitting paper
    • gamma-scale evolution is applied to both isotropic (alpha-iso) and anisotropic (alpha-parallel,alpha-perp) scales
    • Add calculation and printout of zeff for scale measurement
  • 29-Aug-12:

    • Use new generic prior infrastructure in likely package (see flux_Aln2.ini for an example).

    • New config cmds are boxprior[pname] @ (lo,hi), gaussprior[pname] @ (-1sig,+1sig), noprior[pname]
    • Remove old "BAO scale" priors via cmd line options --prior-width, --prior-center.
  • 27-Aug-12:

    • Small changes to CAMB sideband model to match method described in the paper draft.
    • Corresponding small change to CAMB BBand1 models.
    • Allow multiple model-config options that are applied in turn (INI file first).
    • Add EH98 models (see comments in config/flux_Aln2.ini for an example of using these).
  • 22-Aug-12:

    • cmd-line option reuse-cov now takes an integer instead of a boolean.
    • add redshift-evolution for isotropic BAO scale via new "alpha-scale" parameter.
    • protect loadCosmolib(Xi) against invalid indices in .icov input file.
  • 02-Aug-12:

    • Add anisotropic scale parameters to multipole (linearized) and coordinate-space models (full) via new --anisotropic option.
    • Add support for reading French (r,mu) sector files via new ComovingCorrelationData class and --sectors option.

    • New --output-prefix option for running multiple jobs in same directory.
    • Implement mu cuts for finalize, similar to existing r cuts, but only for coordinate space models (new options --mu-min, --mu-max).
    • Add support for saving Fisher information by bin via --save-icov and dumpGradients option to dumpResiduals().
  • 28-Jun-12:

    • Implement MC sampling analysis (--mc-samples, --mc-config).
    • Add cmd-line options to configure/disable prior on BAO scale (--prior-center, --prior-width).
    • Add more options for reading cosmolib Xi files (--xi-rmin, --xi-rmax, --xi-nr, --xi-hexa, --reuse-cov).
  • 24-Jun-12:

    • Implement an optional veto to remove a window around the peak in the fit during the finalize step.
    • Change XiCorrelationModel interpolation to be in r^2 xi(r) instead of xi(r).

    • Rework cmd-line options for XiCorrelationModel (--xi-points, --xi-method).

    • Save a negative error when decorrelated weight is negative in MultipoleCorrelationData.

  • 23-Jun-12:

    • First public release.