Quick unoptimized model to perform certain stat operations

quick_stats.rb

module Statistics
  def self.sum(data)
    data.inject(0.0) { |s, val| s + val }
  end

  def self.mean(data)
    sum(data) / data.size
  end

  def self.median(data)
    data.sort[data.size / 2]
  end

  def self.mode(data)
    val_counts = data.group_by { |v| v }.map { |k, v| [k, v.count] }.sort_by { |k, v| v }.reverse
    val_counts.select { |k, v| v == val_counts[0][1] }.map { |k, _| k }
  end

  def self.stddev(data)
    Math.sqrt(variance(data).abs)
  end

  def self.variance(data)
    mu = mean(data)
    squared_diff = data.map { |val| (mu - val) ** 2 }
    
    sum(squared_diff) / (data.size == 1 ? 1 : data.size - 1)
  end

  def self.dot_product(x, y)
    (0...x.size).inject(0) { |s, i| s + (x[i] * y[i]) }
  end

  def self.linear_reg(x, y)
    m = s_xy(x, y) / s_xx(x)
    b = mean(y) - (m * mean(x))
    [ m, b ]
  end

  def self.s_xy(x, y)
    dot_product(x, y) - sum(x) * sum(y) / x.size
  end

  def self.s_xx(x)
    dot_product(x, x) - sum(x) ** 2 / x.size
  end

  def self.sse(x, y, m, b)
    dot_product(y, y) - m * dot_product(x, y) - b * sum(y)
  end

  def self.sst(y)
    dot_product(y, y) - sum(y) ** 2 / y.size
  end

  def self.r(x, y)
    s_xy(x, y) / (Math.sqrt(s_xx(x)) * Math.sqrt(s_xx(y)))
  end

  def self.r_squared(x, y, m, b)
    1 - sse(x, y, m, b) / sst(y)
  end
end

Small but important change to the variance calculation:

@@ -26,1 +26,1
- sum(data) / (data.size - 1)
+ sum(squared_diff) / (data.size - 1)

This helps to avoid infinite standard deviation in cases where the mean and variance exist in the real number domain.

It's also worth noting that using N-1 (instead of N) is good for finding sample variance and standard deviation, but in the case data.size = 1 (which implies the data is not a sample), unbiased population variance and standard deviation (N) should be used instead to avoid undefined values.

Thanks man. Updated the gist to take both of those into account