# Coefficient of variation to compare running time of two algorithms

I have developed two new sorting techniques using C language. I need to compare the performance of the both sorting techniques, to see which one is better than another. To do this, I used different input data, from n = 500 to n = 2500. For each data that need to be sorted, I run them 10 times and get the mean, standard deviation as well as coefficient of variation. An example of results are shown below:

n          min          max          mean         SD         CV
500        24.07        24.52        21.28        0.11       0.47
1000       52.41        52.83        52.67        0.13       0.25


Just to ask, can I only run each experiment 10 times or should I run more? If I run them 50 times, the results are very close and I obtain almost the same CV. The above results obtained when I run the first sorting technique.

Can I say that 10 times is OK to get the results?

I don't know how many samples should I use to test a particular data. Can anyone suggest any good reading books for this particular analysis?

• Why are you interested specifically in the coefficient of variation? Wouldn't a comparison of the mean execution time be more meaningful? – Gala Jul 24 '13 at 9:40
• which method is the best between the two, one have coeficiency variation of 20% and the other 60% – user37624 Jan 21 '14 at 18:12