Differences in fitting results between PINT and TEMPO2

[weiyu1999]

1、Are the parameters (F0, F1, RA, DEC, etc.) obtained from fitting all pulsars using PINT and TEMPO2 consistent?
2、Under the same data and model conditions, will only a few pulsars yield highly consistent fitted parameters between PINT and TEMPO2?

[dlakaplan]

Hi, @weiyu1999 . Look at section 4 of https://ui.adsabs.harvard.edu/abs/2021ApJ...911...45L/abstract. That has extensive comparison between PINT and tempo(2). Some further testing and comparison is done as part of the PINT CI procedure. It isn't done for all pulsars (that is not possible) but when items such as clocks, ephemerides, and coordinate systems (e.g., there are different possible definitions of ecliptic coordinates) are the same, PINT and tempo(2) give TOA residuals that are consistent to ~30 ns or better. Parameter differences are typically ~0.001 sigma.

[dlakaplan]

Additional tests for the NANOGrav 12.5 year data-set between PINT and tempo are in https://ui.adsabs.harvard.edu/abs/2021ApJS..252....4A/abstract (section 4).

[weiyu1999]

Right Ascension Declination Software
00:00:00.0001(9) +00:00:00.001(4) TEMPO2
00:00:00.0001(3) +00:00:00.001(1) PINT

For example, the above results show the positions fitted using PINT and TEMPO2. The position values are quite close, but the uncertainties differ significantly. However, as mentioned in Section 4 of (https://ui.adsabs.harvard.edu/abs/2021ApJ...911...45L/abstract), the parameter uncertainties for PSR J1600-3053 are nearly identical.

I'm not quite sure why this happens.

[dlakaplan]

Without more detail it's impossible to say what the cause of this difference is. This could be due to using a different fitter type, or scaling uncertainties based on the reduced chi^2. Or something else.

[weiyu1999]

1、How exactly can I adjust the uncertainties based on the reduced chi^2.

2、I couldn't find （docs/examples/J1600-3053_NANOGrav_11yv1.gls.par) and (docs/examples/J1600-3053_NANOGrav_11yv1.tim).
I would like to try fitting it myself.

[dlakaplan]

Depending on how you have PINT installed the example data may be in different places. You can find them at:
https://github.com/nanograv/PINT/tree/master/src/pint/data/examples
And you should be able to find them on your system via the instructions at:
https://nanograv-pint.readthedocs.io/en/latest/tutorials.html#data-files

More generally, PINT does not scale the uncertainties by the reduced chi^2. I think that either TEMPO or TEMPO2 do by default although I cannot remember which (maybe somebody else can remind me). If you look at the reduced chi^2 from your different fits that might help solve this.

[weiyu1999]

Unfortunately, the files J1600-3053.par and J1600-3053.tim mentioned in the paper are not available at (https://github.com/nanograv/PINT/tree/master/src/pint/data/examples).

I don't know where to find them.

[dlakaplan]

You can find the 11yr data release at:
https://nanograv.org/science/data/11-year-pulsar-timing-array-data-release
I am not sure why those files are no longer included in the tests or examples, but there are many other data sets which are.