Stats question
Apr. 8th, 2009 10:47 pmSuppose you have a stochastic algorithm. You run it on 4 instances, a bunch of times each.
Your runtime looks something like this (ignore the y-axis):

You'd like to make a statement about how much of the variability is due to the stochasticity of the algorithm vs the properties of the instances.
What would you compute? Non-parametric ANOVA?
Your runtime looks something like this (ignore the y-axis):
You'd like to make a statement about how much of the variability is due to the stochasticity of the algorithm vs the properties of the instances.
What would you compute? Non-parametric ANOVA?