Modifying P_SN after incoherent time averaging

[JianrongTan]

@jsdillon I have modified the P_SN error bar in single_baseline_postprocessing_and_pspec notebook.

While I am looking at power_spectrum_summary notebook, there is something I am not sure about in the two cells:

# pefrom time average in range set by LST_RANGE_HOURS
lst_subset = uvp.lst_avg_array[(uvp.lst_avg_array >= LST_RANGE_HOURS[0] * np.pi / 12) & (uvp.lst_avg_array <= LST_RANGE_HOURS[1] * np.pi / 12)]
uvp_tavg = uvp.select(lsts=lst_subset, polpairs=[('pI', 'pI')], inplace=False)
uvp_tavg.average_spectra(time_avg=True, error_weights='P_N', error_field=['P_SN'], inplace=True)

In this cell, P_N is not in the error_field.

# correct noise statistics by the coherent average correction factor, which accounts for the FRF
for key in uvp_tavg.get_all_keys():
    uvp_tavg.set_stats('P_N', key, uvp_tavg.get_stats('P_N', key) * dpss_coherent_avg_corrections[key[1]][key[0]]**.5)
    uvp_tavg.set_stats('P_SN', key, uvp_tavg.get_stats('P_SN', key) * dpss_coherent_avg_corrections[key[1]][key[0]]**.5)
    # TODO: Jianrong to figure out whether P_SN should be treated differently here

uvp_tavg seems to have been averaged over interleaves, so there is no such thing like P_S = np.mean([uvp2.get_data(key).real for j, uvp2 in enumerate(uvps_time_avg) if i != j], axis=0) here.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[jsdillon]

On

uvp_tavg.average_spectra(time_avg=True, error_weights='P_N', error_field=['P_SN'], inplace=True)
Would you suggest instead
uvp_tavg.average_spectra(time_avg=True, error_weights='P_N', error_field=['P_SN', 'P_SN'], inplace=True)
?

To get P_S perhaps we should do a temporary de-interleave, then re-interleave? What do you think?

[JianrongTan]

uvp_tavg.average_spectra(time_avg=True, error_weights='P_N', error_field=['P_SN'], inplace=True) Would you suggest instead uvp_tavg.average_spectra(time_avg=True, error_weights='P_N', error_field=['P_SN', 'P_SN'], inplace=True) ?

Yes. I even think we only need error_field=['P_N'] here.

To get P_S perhaps we should do a temporary de-interleave, then re-interleave? What do you think?

Yes. I think we may need a temporary de-interleave and a re-interleave.

[JianrongTan]

There seems no need to change the code for error bar estimation in power_spectrum_summary.ipynb .