Doc changes

RafaelArutjunjan · Aug 5, 2024 · 273e8b6 · 273e8b6
1 parent 2af478c
commit 273e8b6
Show file tree

Hide file tree

Showing 4 changed files with 21 additions and 8 deletions.
diff --git a/docs/src/AdvancedData.md b/docs/src/AdvancedData.md
@@ -2,18 +2,23 @@
 ### Advanced Datasets
 
 
-The following table illustrates the capabilities of the various data types implemented by [**InformationGeometry.jl**](https://github.com/RafaelArutjunjan/InformationGeometry.jl):
+Depending on the nature of the dataset that is analyzed, there are multiple data types implemented by [**InformationGeometry.jl**](https://github.com/RafaelArutjunjan/InformationGeometry.jl) to store them in.
+Mainly, these data types provide a trade-off in speed / simplicity versus flexibility / generality as illustrated by the following table:
 
-| Container                    | allows non-Gaussian `y`-uncertainty | allows `x`-uncertainty | allows mixed `x`-`y` uncertainty | allows missing values |
-|:----------------------------:|:-----------------------------------:|:----------------------:|:--------------------------------:|:---------------------:|
-[`DataSet`](@ref)              |                 ✖                   |           ✖            |                ✖                 |          ✖            |
-[`DataSetExact`](@ref)         |                 ✔                   |           ✔            |                ✖                 |          ✖            |
-[`GeneralizedDataSet`](@ref)   |                 ✔                   |           ✔            |                ✔                 |          ✖            |
-[`CompositeDataSet`](@ref)     |                 ✖                   |           ✔            |                ✖                 |          ✔            |
+
+| Container                      | allows non-Gaussian `y`-uncertainty | allows missing values | allows `x`-uncertainty | allows mixed `x`-`y` uncertainty | allows `y`-uncertainty estimation | allows `x`-uncertainty estimation |
+|:------------------------------:|:-----------------------------------:|:---------------------:|:----------------------:|:--------------------------------:|:---------------------------------:|:---------------------------------:|
+[`DataSet`](@ref)                |                 ❌                  |          ❌           |           ❌           |                ❌                |          ❌                       |          ❌                       |
+[`DataSetExact`](@ref)           |                 ✅                  |          ❌           |           ✅           |                ❌                |          ❌                       |          ❌                       |
+[`CompositeDataSet`](@ref)       |                 ✅                  |          ✅           |           ❌           |                ❌                |          ❌                       |          ❌                       |
+[`GeneralizedDataSet`](@ref)     |                 ✅                  |          ❌           |           ✅           |                ✅                |          ❌                       |          ❌                       |
+[`DataSetUncertain`](@ref)       |                 ❌                  |          ❌           |           ❌           |                ❌                |          ✅                       |          ❌                       |
+[`UnknownVarianceDataSet`](@ref) |                 ❌                  |          ❌           |           ❌           |                ❌                |          ✅                       |          ✅                       |
 
 
 ```@docs
 DataSetExact
 CompositeDataSet
 GeneralizedDataSet
+DataSetUncertain
 ```
diff --git a/docs/src/datamodels.md b/docs/src/datamodels.md
@@ -17,6 +17,8 @@ DataModel
 
 "Simple" `DataSet`s and `DataModel`s can be visualized directly via `plot(DM)` using pre-written recipes for the [**Plots.jl**](https://github.com/JuliaPlots/Plots.jl) package.
 
+
+A more structured container for models is `ModelMap`, which additionally stores information about custom parameter bounds, parameter names, arbitrary nonlinear constraints, etc.:
 ```@docs
 ModelMap
 ```
diff --git a/docs/src/methodlist.md b/docs/src/methodlist.md
@@ -12,6 +12,7 @@ Once a `DataModel` object has been defined, it can subsequently be used to compu
 ```@docs
 loglikelihood(::DataModel,::Vector{Float64})
 MLE(::DataModel)
+MLEuncert(::DataModel)
 LogLikeMLE(::DataModel)
 ```
 
@@ -43,7 +44,7 @@ In many applied settings, one often does not have a dataset of sufficient size f
 A slightly more robust alternative to using geodesics is given by the so-called profile likelihood method. Essentially, it consists of pinning one of the parameters at particular values on a grid, while optimizing the remaining parameters to maximize the likelihood function at every step. Ultimately, one ends up with one-dimensional slices of the parameter manifold along which the likelihood decays most slowly.
 
 ```@docs
-ParameterProfiles(::DataModel,::Int64)
+ParameterProfiles(::DataModel,::Float64)
 InterpolatedProfiles
 ProfileBox
 ```
diff --git a/src/GeneralDataStructures.jl b/src/GeneralDataStructures.jl
@@ -96,6 +96,11 @@ LogPrior(DM::AbstractDataModel) = x::AbstractVector->zero(eltype(x))
 
 xpdim(DM::AbstractDataModel) = Npoints(DM) * xdim(DM) + pdim(DM)
 
+"""
+    MLEuncert(DM::AbstractDataModel, mle::AbstractVector=MLE(DM))
+Returns vector of type `Measurements.Measurement` where the parameter uncertainties are approximated via the diagonal of the inverse Fisher metric.
+That is, the stated uncertainties are a linearized symmetric approximation of the true parameter uncertainties around the MLE.
+"""
 function MLEuncert(DM::AbstractDataModel, mle::AbstractVector=MLE(DM); verbose::Bool=true)
     F = FisherMetric(DM, mle)
     try