'plot_delchi' discrepancies with plot generated using 'get_delchi_plot' arrays

[nplee]

Hi all,

This came in through helpdesk, and I'm able to independently generate a similar case using our standard 3c273.pi example set from the Sherpa threads. It seems like the errors for the x-axis from get_delchi_plot are different than that displayed by plot_delchi.

An example is below:

sherpa In [1]: import matplotlib.pyplot as plt

sherpa In [2]: load_data("3c273.pi")
WARNING: systematic errors were not found in file '3c273.pi'
statistical errors were found in file '3c273.pi' 
but not used; to use them, re-read with use_errors=True
read ARF file 3c273.arf
read RMF file 3c273.rmf
WARNING: systematic errors were not found in file '3c273_bg.pi'
statistical errors were found in file '3c273_bg.pi' 
but not used; to use them, re-read with use_errors=True
read background file 3c273_bg.pi

sherpa In [3]: set_source(powlaw1d.p1)

sherpa In [4]: fit()
WARNING: data set 1 has associated backgrounds, but they have not been subtracted, nor have background models been set
Dataset               = 1
Method                = levmar
Statistic             = chi2gehrels
Initial fit statistic = 8.15838e+10
Final fit statistic   = 50.5351 at function evaluation 25
Data points           = 46
Degrees of freedom    = 44
Probability [Q-value] = 0.231118
Reduced statistic     = 1.14852
Change in statistic   = 8.15838e+10
   p1.gamma       1.72329      +/- 0.0501892   
   p1.ampl        0.000147339  +/- 8.26966e-06 

sherpa In [5]: plot_delchi()

sherpa In [6]: delchiplot = get_delchi_plot()

sherpa In [7]: plt.errorbar(delchiplot.x, delchiplot.y, xerr=delchiplot.xerr, yerr=delchiplot.yerr, ls='none', alpha=0.5)
Out[7]: <ErrorbarContainer object of 3 artists>

sherpa In [8]: print(delchiplot)
x      = [ 0.1245, 0.2774, 0.3869, 0.5183, 0.6059, 0.6716, 0.7227, 0.7665, 0.803 ,
  0.8395, 0.876 , 0.9198, 0.9709, 1.0147, 1.0658, 1.1169, 1.168 , 1.241 ,
  1.3432, 1.4381, 1.5403, 1.6498, 1.752 , 1.8542, 1.9199, 1.9856, 2.0586,
  2.1389, 2.2338, 2.336 , 2.4893, 2.6499, 2.7886, 2.9711, 3.2339, 3.4748,
  3.6792, 3.9128, 4.1391, 4.4822, 4.8691, 5.1976, 5.6356, 6.2342, 8.2198,
 12.41  ]
y      = [-2.781 , 0.2032, 0.9608, 0.0078, 0.7315, 0.4491, 1.3232, 1.1458, 1.3408,
  0.9502, 1.2498,-0.5242, 1.1126, 0.5169,-0.118 , 1.1878, 0.5615,-0.2275,
 -0.4193, 0.9027, 0.3687, 0.7743,-0.2119, 0.1495, 1.5744, 0.0607, 1.1796,
  0.9005, 0.693 , 0.347 ,-0.2612, 1.079 , 0.7726,-0.09  ,-1.2079, 0.8216,
  0.0689,-0.1063, 0.1506,-2.4796,-0.1243,-0.2364,-0.4672,-0.1275,-0.3884,
  3.5365]
yerr   = [1.,1.,1.,1.,1.,1.,1.,1.,1.,1.,1.,1.,1.,1.,1.,1.,1.,1.,1.,1.,1.,1.,1.,1.,
 1.,1.,1.,1.,1.,1.,1.,1.,1.,1.,1.,1.,1.,1.,1.,1.,1.,1.,1.,1.,1.,1.]
xerr   = [0.2467,0.0584,0.1606,0.1022,0.073 ,0.0584,0.0438,0.0438,0.0292,0.0438,
 0.0292,0.0584,0.0438,0.0438,0.0584,0.0438,0.0584,0.0876,0.1168,0.073 ,
 0.1314,0.0876,0.1168,0.0876,0.0438,0.0876,0.0584,0.1022,0.0876,0.1168,
 0.1898,0.1314,0.146 ,0.219 ,0.3066,0.1752,0.2336,0.2336,0.219 ,0.4672,
 0.3066,0.3504,0.5256,0.6716,3.2996,5.0808]
xlabel = Energy (keV)
ylabel = Sigma
title  = Sigma Residuals for 3c273.pi
plot_prefs = {'xerrorbars': True, 'yerrorbars': True, 'ecolor': None, 'capsize': 0, 'barsabove': False, 'xlog': False, 'ylog': False, 'linestyle': 'None', 'drawstyle': 'default', 'color': None, 'marker': '.', 'markerfacecolor': None, 'markersize': None, 'alpha': None, 'xaxis': True, 'ratioline': False, 'linecolor': None}

where in the figure, the black figure is from plot_delchi and the translucent red figure is the overplotted pyplot using the get_delchi_plot arrays. From what I understand, the two functions are dependent on the DelchiPlot class:

https://github.com/sherpa/sherpa/blob/767f516b929f7f53ae256a4902edd27fec4303b4/sherpa/plot/init.py#L1957-L2009

The x-errors are the +/- half-width of the energy/wavelength bins and the y-errors are just +/-1; and I'm not convinced that the error bars in the delchi plot are particularly meaningful.

Anyways, the x-error bar array plotted by pyplot.errorbar are larger than that plotted by by plot_delchi even though it doesn't look like the Sherpa function is manipulating the values... from mere speculation makes me wonder if something's going on with how grouping is being handled?