Takes CSV. Calculates pairwise correlations (excludes empty cells)

correlation-inator.rb

require 'csv'

# correlation-inator.rb
# by Charles (c650)
# Takes in CSV, calculates pairwise correlations but excluding empty cells

class Pair
	attr_accessor :a, :b
	def initialize(a, b)
		a.gsub!(/#/,"")
		b.gsub!(/#/,"")
		@a = a.to_f
		@b = b.to_f
	end
end

if ARGV.empty?
	puts "Usage: ./doit.rb [path to csv]"
	exit
end

data = CSV.read(ARGV[0])

correl_matrix = Array.new(data.length-1){Array.new(data.length-1, 1000)}

point_names = [nil]

for i in 1...data.length
	point_names << data[i][0]
	for j in (i+1)...data.length
		tmp = []
		for k in 1...[data[i].length, data[j].length].min
			tmp << Pair.new(data[i][k], data[j][k]) unless data[i][k].nil? || data[j][k].nil?
		end
		x_sum = 0
		y_sum = 0

		tmp.each do |pair|
			x_sum += pair.a;
			y_sum += pair.b;
		end

		x_avg = x_sum / tmp.length
		y_avg = y_sum / tmp.length

		top_sum = 0
		x_bottom_sum = 0
		y_bottom_sum = 0

		tmp.each do |pair|
			top_sum += (pair.a - x_avg) * (pair.b - y_avg)
			x_bottom_sum += (pair.a - x_avg) ** 2
			y_bottom_sum += (pair.b - y_avg) ** 2
		end

		correl_matrix[i-1][j-1] = correl_matrix[j-1][i-1] = top_sum / Math.sqrt(x_bottom_sum * y_bottom_sum)
	end
end

CSV.open("./out.csv", "wb") do |csv|
	csv << point_names
	correl_matrix.each_with_index do |row, i|
		tmp = [point_names[i+1]]
		tmp << row
		tmp.flatten!
		csv << tmp
	end
end