Create Unified Hierarchy of Models and their names

[aaronpeikert]

related #32 #82

[LeonieHagitte]

We thought of the following structure:

Taxon (lvl1)
pathmodels (#relations btw. manif. variables)(lvl2) cfa (#relations of lat onto manif. var, which are not correlated) general sem (#combination of both) (lvl2) simple cfa( #as part of cfa #81 #standalone factor) (lvl3) hierarch. cfa( #as part of cfa) (lvl3)
Longit. mods (#as oart of general sem) (lvl3) cross-sect. models(#a part of general sem) (lvl3)

[LeonieHagitte]

Whether it is important to differentiate between hierarchical and non hierarchical cfas, we are uncertain. It would also be possible to differentiate only via taxons in the struct etc./ new labels for the secondorder factors etc. What is the general opinion on this?

[aaronpeikert]

Can you write it as:

abstract type Taxon end
struct PathModel <: Taxon end
abstract type SEM <: Taxon end
struct CFA <: SEM end

?

[aaronpeikert]

Whether it is important to differentiate between hierarchical and non hierarchical cfas, we are uncertain. It would also be possible to differentiate only via taxons in the struct etc./ new labels for the secondorder factors etc. What is the general opinion on this?

I think it makes sense to differentiate CFA types. Consider Hierachical vs Bifactor:

[aaronpeikert]

Also consider: https://openmx.ssri.psu.edu/wiki/modeltypes

[valentinkm]

[aaronpeikert]

Very good! Now, change the Taxonomy package in a PR so that:

using Taxonomy, AbstractTrees
AbstractTrees.children(d::DataType) = subtypes(d)
print_tree(Taxonomy.Taxon)
Taxon
├─ LGCM
├─ NoTaxon
├─ AbstractFactor
│  └─ Factor
└─ Pathmodel
   └─ CFA

is representing this structure.

[aaronpeikert]

Easy right :rofl:

[LeonieHagitte]

'abstract type Taxon end abstract type NoTaxon end struct PathModel <: Taxon end struct CFA <: Taxon end struct General_SEM <: Taxon end struct simple_CFA <: CFA end struct hierarchical_CFA <: CFA end struct longitudinal_SEM <: General_SEM end struct crosssectional_SEM <: General_SEM end'

[aaronpeikert]

Learn about CamelCase:

https://docs.julialang.org/en/v1/manual/style-guide/#Use-naming-conventions-consistent-with-Julia-base/

Learn about abstract vs concrete:

abstract type IAmAbstract end
struct IamConcrete end

# works:

struct SubType <: IAmAbstract end

# doesn't work:

struct SubType <: IAmConcrete end

https://docs.julialang.org/en/v1/manual/types/#man-abstract-types

[LeonieHagitte]

okay this changes things :)

abstract type Taxon end abstract type NoTaxon end struct PathModel <: Taxon end abstract type CFA <: Taxon end abstract type General_SEM <: Taxon end struct simple_CFA <: CFA end struct hierarchical_CFA <: CFA end struct longitudinal_SEM <: General_SEM end struct crosssectional_SEM <: General_SEM end'

[aaronpeikert]

Great, now only CamelCase: https://docs.julialang.org/en/v1/manual/style-guide/#Use-naming-conventions-consistent-with-Julia-base/ e.g.: hierarchical_CFA -> HierachicalCFA

Also, I suggest renaming General_SEM to AbstractSEM.

[aaronpeikert]

You may take insperation from other packages regarding naming, e.g., this is how stenograph looks:

julia> print_tree(StenoGraphs.AbstractEdge)
AbstractEdge
├─ Edge
│  ├─ DirectedEdge
│  └─ UndirectedEdge
└─ MetaEdge
   ├─ Arrow
   └─ ModifiedEdge

julia> print_tree(StenoGraphs.AbstractNode)
AbstractNode
├─ MetaNode
│  ├─ ModifiedNode
│  └─ ModifyingNode
└─ Node
   └─ SimpleNode