Concepts DL4:

[kosack]

We discussed that the structure of Level2 (DL4) can be similar to Level 1 (DL3), except that a collection of Observations → DataSet

classDiagram

    L1.Observation "1..*" --> "1" L2.DataSetCollection 
    L2.DataSetCollection -- "1..*" L2.DataSet
    L2.DataSet -- L2.IRF
    L2.IRF -- ExposureMap
    L2.IRF -- PSFMap
    L2.IRF -- EDispMap
    L2.IRF -- BkgMap
   
    L2.DataSet -- L2.MaskMap

    L2.DataSet -- L2.Events
    L2.Events -- "0..1" L2.CountsMap
    L2.Events -- "0..1" L2.EventList

Loading

L2.EventList is e.g. for when you have stacked a bunch of IRFs, and maybe made an event selection, and therefore the original L1.EventLists are not sufficient. So being able to optionally store a new

SkyModels? Suggestion is to just put them into L3. That avoids ambiguity with input/output/statics.

[kosack]

Naming?

DataSets should have a "value_ucd" that can be e.g.

• L2 = "InstrumentDataSet", value_ucd = "counts" (or other instrumental units)
• L3 = "PhysicsDataSet", value_ucd = "flux" (or other physics units)

Note that the coordinates of the spatial dimensions of the maps in L2/L3 are both in "sky" units, not detector units.

[kosack]

Do we need two types?

classDiagram
 
   L2.StackedDataSet --|> L2.DataSet
   L2.JointDataSet --|> L2.DataSet

Loading

Actually, a JointDataSet is just a list of DataSets, so it's equivalent to the ObservationCollection

So really we have:

classDiagram
 
   L1.Observation -- L2.DataSetCollection 
   L2.DataSetCollection -- "1..*" L2.DataSet

Loading

[kosack]

Maps:

classDiagram
    class Map {
           + geometry
    }
    Map *-- Value
    Map *-- "0..1" ValidityMask

Loading