MaNGA Caveats

The primary source code for the Data Reduction Pipeline (DRP) is IDL, which a row-major language. When using a column-major language, like Python, it is important to understand the difference in the ordering of the datacube. For example, the ‘FLUX’ extension in the DRP datacube fits files are written by IDL with the three dimensions organized as (RA, DEC, wavelength); however, when read in by, e.g., astropy.io.fits, the array order is inverted to (wavelength, DEC, RA). Please see the MaNGA Python Tutorial for example code.

Each MaNGA data cube has an associated 3d maskbit array describing the quality of a given value DRP3PIXMASK , and whether it should be used in an analysis. This includes effects such as the IFU footprint, missing data, foreground stars (where known), etc. Any use of the MaNGA data should consider these maskbits.

DR13 does not provide covariance calculations for the provided datacubes; however, there is significant covariance between adjacent spaxels. When combining spectra from multiple spaxels, a rigorous calculation of the inverse variance in the combined spectrum must account for this covariance. Short of that, we provide a calibration of the noise vector calculated without considering the covariance to a calculation that does. The calibration is:

n_covar/n_{no covar} = 1 + 1.62 log₁₀(N_bin),

for N_bin ≤ 100
and

n_covar/n_{no covar} = 4.2,

for N_bin > 100

where n_{no covar} is determined via a nominal error calculation using the inverse variance provided in the datacube and N_bin is the number of binned spaxels. The correction factor is constant above N_bin = 100 because additional spaxels at that point are uncorrelated with the original spaxels. It is important to note that this calibration is dependent on the spaxels being adjacent.

See further discussion in Law et al. (2016) (preprint available here)

Early analysis comparing MaNGA galaxies to high-resolution DiskMass (Bershady et al. 2010) observations suggests that the spectral resolution of the DR13 MaNGA data is overestimated by about 10% (i.e., the instrumental line spread function is about 10% larger than reported by the DRP data products). This affects both per-fiber estimates of the line spread function (e.g., the DISP extension of the summary RSS files) and the summary average estimates of the spectral resolution (e.g., the SPECRES extension of the data cubes). This discrepancy is due to the combined effects of smoothing introduced by the wavelength rectification step, and the difference between gaussian width measurements assuming integration of the gaussian profile over the pixel boundaries vs evaluation of the profile at the pixel midpoint, both of which will be addressed in a future DR. We note that this effect is common to all prior SDSS optical spectra, not just MaNGA.

See further discussion in Law et al. (2016) (preprint available here).

Although most cosmic rays and other transient features are detected by the DRP and flagged (either for removal or for masking), lower-intensity glitches can make it into the final datacubes and show up as occasional hot pixels. Further improvements to the DRP to handle these spurious cosmic rays is still under development. Currently, some of these artifacts are being addressed on a case-by-case basis; however, individual hot glitches are not addressed. It is particularly important to be wary of this when searching for isolated emission features in the data cubes.

The 3D phase of the DRP has an overall reduction quality bit DRP3QUAL that indicates any potential quality control issues with a given output file for each observation. Most of these issues, like shallow observations, are simply warnings that the data might not be of the usual quality. Flux-calibration failures, however, trigger the CRITICAL quality bit, which indicates that there may be severe problems with the data. This is determined by whether or not the astrometric calibration is successful without a substantial rescaling of the flux to match the imaging data.

Critical failures occur in roughly 2% of observations. These are a mixture of true critical failures (where, e.g., an IFU is badly out of focus, such as 7495-6103) and less critical issues where transients or bright objects at the edges of the field cause problems with the astrometric solution. Reasons for the latter can include some instances where the on-sky surface brightness distribution seems to be genuinely different from that predicted by the preimaging (such as 8332-12702). In some cases the extra flux comes from a terrestrial transient (satellite trails, etc). Further development is underway to address these issues.

The Secondary sample was designed to have a higher density of targets than is required to give the desired 2:1 ratio of Primary+ to Secondary galaxies. Therefore, before allocating IFUs we random sample the Secondary sample to the desired target density. Due to a very recently discovered bug in the target selection code this random sampling is not truly random and in fact samples in such a way as to make the number density distribution flat as a function of stellar mass. This is a small change, since the density distribution was already quite close to being flat with stellar mass. However, it does mean the observed Secondary sample is no longer selected purely by i-band absolute magnitude and redshift, but also has a weak dependence on stellar mass. It also has consequences for calculating the appropriate weights to apply to any sample containing the Secondary sample (see the weights FAQ). At the moment the information required to calculate the weights taking this bug into account is not available within DR13 but we plan to make it available soon. In the mean time ignoring this weak mass dependence in sampling probability of the Secondary sample and just using the weights as described in the weights FAQ should be sufficiently accurate for most purposes. Systematic error due to this effect should be smaller than the sample variance with the current sample size.