Merge 77b7810 into 9b49f98

src/regression_sensitivity.jl

@@ -0,0 +1,49 @@
+using RecursiveArrayTools, DataFrames, GLM
+function regression_sensitivity(f,p_range,p_fixed,n)
+    pearson_coeffs = []
+    src_coeffs = []
+    for i in 1:length(p_range)
+        pcs = []
+        srcs = []
+        params = []
+        for j in 1:length(p_range)
+            if i != j
+                push!(params,p_fixed[j])
+            else
+                push!(params,0)
+            end
+        end
+        xis = []
+        yis = Array{Float64}[]
+        for j in 1:n
+            params[i] = (p_range[i][2]-p_range[i][1])*rand() + p_range[i][1]
+            x = params
+            y = f(params)
+            push!(xis,copy(x))
+            push!(yis,copy(y))
+        end
+        yis = VectorOfArray(yis)
+        x_mean = mean(xis)
+        x_vrs = [sum(xis[i] - x_mean) for i in 1:length(xis)]
+        x_lm = [j[i] for j in xis]
+        x_var = var(x_lm)
+        for k in 1:size(yis)[1]
+            pc = []
+            src = []
+            for j in 2:size(yis)[2]
+                pear_coeff_num = sum(x_vrs .* (yis[k,j,:] - mean(yis[k,j,:])))
+                pear_coeff_deno = sqrt(sum(x_vrs.^2)) * sqrt(sum((yis[k,j,:] - mean(yis[k,j,:]).^2)))
+                pear_coeff = pear_coeff_num/pear_coeff_deno
+                push!(pc,pear_coeff)
+                df = DataFrame(X=x_lm,Y=yis[k,j,:])
+                lin_model = lm(@formula(Y ~ X),df)
+                push!(src,coef(lin_model)[2].*sqrt.(x_var ./ var(yis[k,j,:])))
+            end
+            push!(pcs,pc)
+            push!(srcs,src)
+        end
+        push!(pearson_coeffs,pcs)
+        push!(src_coeffs,srcs)
+    end
+    src_coeffs
+end