Rel 0.1.0

Author: goedman

README.md

@@ -10,6 +10,7 @@
 [![codecov.io](http://codecov.io/github/StatisticalRethinkingJulia/StructuralCausalModels.jl/coverage.svg?branch=master)](http://codecov.io/github/StatisticalRethinkingJulia/StructuralCausalModels.jl?branch=master)
 [![Documentation](https://img.shields.io/badge/docs-stable-blue.svg)](https://StatisticalRethinkingJulia.github.io/StructuralCausalModels.jl/stable)
 [![Documentation](https://img.shields.io/badge/docs-master-blue.svg)](https://StatisticalRethinkingJulia.github.io/StructuralCausalModels.jl/dev)
+[![ColPrac: Contributor's Guide on Collaborative Practices for Community Packages](https://img.shields.io/badge/ColPrac-Contributor's%20Guide-blueviolet)](https://github.com/SciML/ColPrac)
 
 # Introduction
 

docs/src/index.md

@@ -16,7 +16,7 @@ DAG(name::AbstractString, d::OrderedDict, df::DataFrame)
 ## d_separation
 ```@docs
 d_separation(d::DAG, first::SymbolList, second::SymbolList; debug=false)
-```
+d_separation(d::DAG, first::SymbolList, second::SymbolList, cond::SymbolList; debug=false)```
 
 ## shipley_test
 ```@docs

docs/src/walkthrough.md

@@ -77,7 +77,7 @@ Rows ╲ Cols │  :mechanics     :vectors     :algebra    :analysis  :statistic
 
 Additional DAG related functions are `adjacency_matrix()`, `edge_matrix()`, `to_ggm()`, `from_ggm()`, `to_dagitty()`, `from_dagitty()`, `set_dag_df!()` and `set_dag_cov_matrix!()`.
 
-# D_separation
+# Directed separation
 
 Given a causal graph, `d_separation(dag, f, l, cond)` determines if the vertices in set `f` are `d-separated` from the vertices in set `l` given the conditioning set `cond`.
 
@@ -194,10 +194,9 @@ fname = scm_path("..", "examples", "SR", "SR6.4.3", "sr6.4.3.dot")
 Sys.isapple() && run(`open -a GraphViz.app $(fname)`)
 
 d = OrderedDict(
-  :s => [:a, :m, :w],
-  :a => [:m, :d],
-  :m => [:d],
-  :w => [:d]
+  [:a, :m, :w] => :s,
+  :d => [:w, :m, :a],
+  :m => [:a]
 );
 u = []
 
@@ -225,7 +224,7 @@ println()
 ```
 
 for this [DAG](https://github.com/StatisticalRethinkingJulia/StructuralCausalModels.jl/blob/master/examples/SR/SR6.4.3/sr6.4.3.pdf) returns:
-```
+```julia
 4-element Array{Array{Symbol,1},1}:
  [:w, :s, :a, :d]
  [:w, :s, :a, :m, :d]
@@ -245,35 +244,15 @@ println()
  [:w, :s, :m, :d]
  [:w, :s, :m, :a, :d]
 
-openpaths = open_paths(dag, backdoorpaths)
-println("All open (backdoor) paths between :w and :d:")
-openpaths |> display
-println()
-
-3-element Array{Array{Symbol,1},1}:
- [:w, :s, :a, :d]
- [:w, :s, :a, :m, :d]
- [:w, :s, :m, :d]
 ```
-
-Not exported `blocking_sets()`:
-```jula
-StructuralCausalModels.blocking_sets(openpaths)
-4-element Array{Array{Symbol,1},1}:
- [:w]
- [:s]
- [:d]
- [:a, :m]
-
+```julia
 println("Show path: $(allpaths[2])")
 show_dag_path(dag, allpaths[2]) |> display
 println()
 
 ":w ⇐ :s ⇒ :a ⇒ :m ⇒ :d"
 ```
 
-The vertices :w and :d will be removed by `adjustment_sets()`.
-
 # Ancestral graph
 
 Setup an ancestral_graph example:

examples/SR/SR6.4.3/sr6.4.3.jl

@@ -21,10 +21,9 @@ if isdefined(Main, :StatsPlots)
 end
 
 d = OrderedDict(
-  :w => :s,
+  [:w, :m, :a] => :s,
   :d => [:a, :w, :m],
-  :m => [:a, :s],
-  :a => :s
+  :m => [:a]
 );
 
 dag = DAG("sr6_4_3", d, df);

src/StructuralCausalModels.jl

@@ -42,6 +42,7 @@ include("methods/basis_set.jl")
 include("methods/shipley_test.jl")
 include("methods/d_separation.jl")
 include("methods/pcor.jl")
+include("methods/pcor_test.jl")
 include("methods/induced_covariance_graph.jl")
 include("methods/all_paths.jl")
 include("methods/open_paths.jl")

src/methods/d_separation.jl

@@ -18,7 +18,7 @@ d_separation(
 
 ### Optional arguments
 ```julia
-* `cond::Vector{Symbol}`               : Conditioning set
+* `cond::SymbolList`                   : Conditioning set
 * `debug=false`                        : Trace execution
 ```
 
@@ -61,17 +61,16 @@ The Julia translation is licenced under: MIT.
 
 Part of the API, exported.
 """
-function d_separation(d::DAG, first::SymbolList, second::SymbolList,
-  cond::SymbolList; debug=false)
+function d_separation(d::DAG, first::SymbolList, second::SymbolList; debug=false)
 
-  e = induced_covariance_graph(d, vcat(first, second), cond; debug=debug)
+  e = induced_covariance_graph(d, vcat(first, second), SymbolList[]; debug=debug)
   sum(e[first, second]) == 0
 
 end
 
-function d_separation(d::DAG, first::SymbolList, second::SymbolList; debug=false)
+function d_separation(d::DAG, first::SymbolList, second::SymbolList, cond::SymbolList; debug=false)
 
-  e = induced_covariance_graph(d, vcat(first, second), SymbolList[]; debug=debug)
+  e = induced_covariance_graph(d, vcat(first, second), cond; debug=debug)
   sum(e[first, second]) == 0
 
 end

src/methods/pcor.jl

@@ -58,3 +58,6 @@ function pcor(u::Vector{Symbol}, S::NamedArray)
   k = inv(S[us, us])
   -k[1,2] / sqrt(k[1,1] * k[2,2])
 end
+
+export
+  pcor

src/methods/pcor_test.jl

@@ -0,0 +1,61 @@
+#=
+"pcor.test" <-
+function(r, q, n){
+                df = n - 2 - q
+                tval <- r * sqrt(df)/sqrt(1-r*r)
+                pv <- 2 * pt(-abs(tval), df)
+  list(tval = tval, df = df, pvalue = pv)
+
+}
+=#
+
+"""
+    
+# pcor
+
+$(SIGNATURES)
+
+Computes the partial correlation between two variables given a set of other
+variables.
+
+### Method
+```julia
+pcor_test(;
+* `u::Vector{Symbol}`                  : Variables used to compute correlation
+* `S::Matrix`                          : Sample covariance matrix
+)
+```
+where:
+
+  u[1], u[2]: Variables used to compute correlation between, remaining variables
+  are the conditioning set
+
+
+### Returns
+```julia
+* `res::Float64`                       : Correlation between u[1] and u[2]
+```
+# Extended help
+
+### Acknowledgements
+
+Original author:                       Giovanni M. Marchetti
+
+Translated to Julia:                   Rob J Goedman
+
+### License
+
+The R package ggm is licensed under License: GPL-2.
+
+The Julia translation is licenced under: MIT.
+
+Part of the api, not exported.
+"""
+function pcor_test(r, q, n)
+
+  df = n - 2 - q
+  tval = r * sqrt(df) / sqrt(1 - r*r)
+  pv = 2 * pdf(TDist(-abs(tval), df))
+  (pv=tval, df=df, pvalue=pv)
+
+end