views:

310

answers:

2

I've written two simple calculations with Ruby which match the way that Microsoft Excel calculates the upper and lower quartiles for a given set of data - which is not the same as the generally accepted method (surprise).

My question is - how much and how best can these methods be refactored for maximum DRYness?

# Return an upper quartile value on the same basis as Microsoft Excel (Freund+Perles method)
  def excel_upper_quartile(array)
      return nil if array.empty?
      sorted_array = array.sort
      u = (0.25*(3*sorted_array.length+1))
      if (u-u.truncate).is_a?(Integer)
        return sorted_array[(u-u.truncate)-1]
      else
        sample = sorted_array[u.truncate.abs-1]
        sample1 = sorted_array[(u.truncate.abs)]
        return sample+((sample1-sample)*(u-u.truncate))
      end
  end


  # Return a lower quartile value on the same basis as Microsoft Excel (Freund+Perles method)
  def excel_lower_quartile(array)
      return nil if array.empty?
      sorted_array = array.sort
      u = (0.25*(sorted_array.length+3))
      if (u-u.truncate).is_a?(Integer)
        return sorted_array[(u-u.truncate)-1]
      else
        sample = sorted_array[u.truncate.abs-1]
        sample1 = sorted_array[(u.truncate.abs)]
        return sample+((sample1-sample)*(u-u.truncate))
      end
  end
+1  A: 

Some might disagree on the refactoring, but here's how I'd handle it:

def excel_quartile(extreme,array)      
  return nil if array.empty?
  sorted_array = array.sort
  u = case extreme
  when :upper then 3 * sorted_array.length + 1
  when :lower then sorted_array.length + 3
  else raise "ArgumentError"
  end
  u *= 0.25
  if (u-u.truncate).is_a?(Integer)
    return sorted_array[(u-u.truncate)-1]
  else
    sample = sorted_array[u.truncate.abs-1]
    sample1 = sorted_array[(u.truncate.abs)]
    return sample+((sample1-sample)*(u-u.truncate))
  end
end

def excel_upper_quartile(array)
  excel_quartile(:upper, array)
end

def excel_lower_quartile(array)
  excel_quartile(:lower, array)
end
EmFi
+1  A: 

I'll start by generalizing a little and provide one method to handle both instances.

def excel_quartile(array, quartile)
  # Returns nil if array is empty and covers the case of array.length == 1
  return array.first if array.length <= 1
  sorted = array.sort
  # The 4th quartile is always the last element in the sorted list.
  return sorted.last if quartile == 4
  # Source: http://mathworld.wolfram.com/Quartile.html
  quartile_position = 0.25 * (quartile*sorted.length + 4 - quartile)
  quartile_int = quartile_position.to_i
  lower = sorted[quartile_int - 1]
  upper = sorted[quartile_int]
  lower + (upper - lower) * (quartile_position - quartile_int)
end

Then you can make convenience methods of:

def excel_lower_quartile(array)
  excel_quartile(array, 1)
end

def excel_upper_quartile(array)
  excel_quartile(array, 3)
end

Note: the excel_quartile method matches expectations for quartile in { 1, 2, 3, 4}. Anything else, I guarantee failure.

Update:

The formula I used is not expressly given at the website I cited, but it is the abstraction for the Freund and Perles method of calculating the quartile position.

Further update:

There is an error in your original code, though you should never encounter it: u - u.trunc is always within the interval [0.0, 1.0), thus the only time it would resemble an integer is when u - u.trunc = 0. However, (u - u.trunc) is still an instance of a Float whenever u is a Float, so your code never happens upon the miscalculated index. Incidentally, if u - u.trunc were an integer, your method would return the last element of the array.

Ian