I have different implementations of functions in a computer program.
Now, I want to find out if one of the implementations shows an increase in performance.
To do that, I implemented a slow function and 2 faster versions and measured the run-time for each. The problem is, that the recorded samples have different distributions.
For example:
The sample of the slow function $X$ is normal distributed, whereas the sample of a faster version $Y$ is not (Gamma- or F like distribution, sometimes equal distributed ).
I want to show that: $\bar{Y} < \bar{X}$
By calculating the mean run-times, I can see that the faster version $Y$ provides an improvement. But how can I use statistical tests to show or prove this?
I only know the t-test, z-test but they assume t- or normal distributed samples which is not the case for my data.
Thank you for your help :)