Data Challenge 1 has a very simplistic representation of IPTA data. Here is a short list of some of the ways that real data differ from the data challenge 1 data. In no particular order:
- TOA uncertainties are different for each observation. Some pulsars exhibit large changes in signal-to-noise ratio due to scintillation in the interstellar medium. This dramatically changes the measurement uncertainty from day to day.
- Gaps exist in the data. Telescopes go down for maintenance, observers make a mistake. Some of these gaps could be months or more, depending on the pulsar and telescope.
- TOA uncertainties may not reflect the real “white noise” level. The uncertainties in TOA measurement may be over or under estimated. This may be due correctable in future, but currently there seem to be sources of noise outside our control/understanding.
- TOA uncertainties change with time. Due to constant improvement in instrumentation, the average uncertainty typically decreases as a function of time.
- Different telescopes have different properties. Different telescopes in the IPTA have different properties. Some pulsars have data from multiple telescopes. Different PTAs have different cadences, different observing frequencies, different algorithms for computing arrival times, different ways of correcting instrumental delays, different profile templates, etc. This means that the statistics of the datasets from different PTAs is likely to be different.
- The ISM delay varies as a function of time. The interstellar medium introduces an observing frequency dependant delay in the residuals. This changes with time. In principal this can be corrected using multi-frequency observations, however this is not trivial and is not included in the simulations. There may also other delays that are not typically accounted for, depending on the pulsar.
- Archival data is of much lower quality. It is likely that a final search for the GWB on real data will use archival data for the IPTA pulsars. This often has a much worse quality, and may not have multiple frequencies meaning that DM correction cannot be done. This means that the statistics will be different for the archival data to the “IPTA” data.
- Pulsars have different data spans. Some pulsars have been added to the array, some have been removed. Archival data on IPTA pulsars varies dramatically.
- Pulsars have unmodelled red noise, with unknown spectrum and amplitude. This is hard to quantify, since we do not know how much may be due to the GWB, but it appears that many pulsars have red noise. Of particular note is PSR J1939+2134, which has a very large cubic signal in the data. This is very unlikely to be from the GWB since it does not appear in other pulsars. Our simulations do include some simple model for red noise, however it is possible that reality will be quite different.