0

I have an array A where each element is an Mean Squared Error. How can I calculate the mean of A?

If I do a simply mean (If I do so I should got a mean of means) of the elements of A, is it a correct operation? If not why? And what's a solution?

Note: The elements in A are real in range from 0 to 1.

Ewybe
  • 395
  • 2
  • 4
  • 15
  • Sounds like more of a theoretical maths question, in which case it should be move to http://math.stackexchange.com. But my instinct is yes, just take a mean; what properties of the result do you need that that wouldn't fulfil? – IMSoP Jun 28 '14 at 12:38

1 Answers1

0

If you're after the total mean squared error you'll need the number of values that contributed to each element, n[i][j]. You can then compute

total_err2 = (Σ (n[i][j] * err2[i][j])) / (Σ n[i][j])

where Σ is the sum over all of the elements.

thus spake a.k.
  • 1,607
  • 12
  • 12