WeightMixer: hybrid cross-power spectrum estimation
Lastest updated Planck-targetted code (e.g. with polarization) now included in LensPix.
This page gives information about the older code specifically for analysing WMAP temperature maps.
The code requires HEALPix 2.x and allows you to:
The code includes all of LensPix - the MPI framework used for doing spherical transforms (based on HealPix).
- Generate power spectrum estimators from WMAP 5-year maps
- Generate hybrid cross- and auto- power spectrum and covariance from general foreground-cleaned maps
- Simulate combined maps or combinations of maps for individual detectors
- Do MPI spherical transforms of arrays of maps, calculate coupling matrices etc.
If you use the code you should of course remember to cite HEALPix as instructed on their page and the relevant WMAP papers.
- Does not include beam uncertainties (or position dependence, time dependence, etc.)
- All noise is assumed to be Gaussian and uncorrelated between pixels, detectors and years.
- Pixels are assumed to be exact averages of the CMB over the pixel (plus noise).
- Polarization support limited (i.e. single maps only and noise proprtional to temperature noise)
See the appendix of arXiv:0801.0554. Untested in this code.
- There is quite a lot of undocumented/legacy code to do other things (e.g. large scale harmonic likelihood)
- Not fully MPI - many things only done on main thread.
- Download the code
- Compile the code in the source directory by running Make (after editing the Makefile for your file locations; code is currently tested only with ifort compiler).
- Edit params.ini for your run settings
- To use the default settings to analyse WMAP5 you need to download and extract to the same directory somewhere:
- To run the code use e.g. mpirun -np 8 weightmixer params.ini. You may also be able to adapt the supplied job script.
The supplied parameter file will generate the hybrid mixing matrix, approximate estimator covariance, power spectrum and combined point source spectrum. It takes about 20 minutes for the first run on 8 processors. Subsequent runs are faster as things are cached in the data/ directory: you may need to delete files in this directory if you don't want them to be cached. Output power spectra etc. are int the outfiles directory. To analyse to simulate other data you would need an equivalent set of input maps, beam transfer functions, etc. in a consistent format.
The WMAP analysis is discussed in arXiv:0804.3865. The polarization analysis in the appendix of arXiv:0801.0554.
You can download the spectrum, noise and point source data from the paper. The files are in the format used by the latest version of the Planck_like likelihood module for CosmoMC.
- Neff file output includes point source spectrum at fiducial value
- Does not consistently deal with monopole and dipole
- Should add regularization of hybrid at low l
Antony Lewis April 2008.