# Measuring statistical consistency and reliability

I am trying to score reliability (or variability) for a machine learning model I'm working on and I cannot quite get my head around it. As an example, I have some people who perform a task three times per week for 10 weeks. I have other people who perform the same task, say, 10 times for three weeks during the 10 week period. And maybe I have people performing the task 30 times in one week and resting over the other nine weeks.

They all average 3 tasks per week, but I need to give them different variabilities. How would I score that?

• Do you want a measure of how "spread out" they work during the 10 weeks? So for instance performing the task 30 in one week maps to $1$, while performing three tasks per week for 10 weeks maps to $0$ (if these are the extreme cases) and 10 times for three weeks maps to something in between? Commented Oct 1, 2016 at 18:42
• Yes... that sounds right. I want to understand who performs better if they do it all at once or spread it out. Commented Oct 1, 2016 at 18:56

Note that it doesn't matter where the $10$s (or the $30$) are placed in the list.