Lightcurve new features

[henrikef]

Diagnostic plots

• [ ] Exposure vs time
• [ ] nPred vs time
• [ ] relative uncertainty vs TS
• [ ] Distribution of flux, uncertainty, index, ...

Other features

• [ ] Make convergence strategy (and fallback strategy) user configurable
• [ ] Allow to select sources by variability TS
• [ ] Different convergence strategies (adjust fit precision)
• [ ] Improve support for parallelization: ipyparallel(?)
• [ ] Put failed bins at np.nan
• [ ] Keep status flag about which step in the strategy was used
• [ ] Export flux likelihood profile per bin (if not already there)

[adonath]

Maybe this is of interest here: in Gammapy we have now a generalized FluxPoints class, which stores an n-dimensional set of flux points. Internally it uses the exact same representation as fermipy, based on a reference model and an array of "norm" values. In general this means flux points can vary with arbitrary physical axes, such as energy, time, spatial coordinates or e.g. event types. This makes plotting certain quantities very simple, see e.g.

• https://docs.gammapy.org/1.0/tutorials/analysis-time/light_curve.html#night-wise-lc-estimation
• https://docs.gammapy.org/1.0/tutorials/analysis-3d/flux_profiles.html#plotting-results
• https://docs.gammapy.org/1.0/tutorials/analysis-time/light_curve_simulation.html#extract-the-lightcurve
• https://docs.gammapy.org/1.0/api/gammapy.estimators.FluxPoints.html#gammapy.estimators.FluxPoints

As fermipy already uses Gammapy as a dependency, this might save some work here.