statistics - How do I measure the distribution of an attribute of a given population? -


i have catalog of 900 applications. need determine how reliability distributed whole. (i.e. normal).

i can measure reliability of individual application.

how can determine reliability of group whole without measuring each one?

that's pretty open-ended question! overall, distribution fitting can quite challenging , works best large samples (100's or 1000's). it's better pick modeling distribution based on known characteristics of process you're attempting model try purely empirical fitting.

if you're going go empirical, start take random sample, measure reliability scores (whatever you're using that) of sample, sort them, , plot them vs normal quantiles. if fall along relatively straight line normal distribution plausible model, , can estimate sample mean , variance parameterize it. can apply same idea of plotting vs quantiles other proposed distributions see if plausible well.

watch out behavior in tails, in particular. pretty definition tails occur , may under-represented in sample. things statistical, larger sample size can draw on better results be.

i'd add prior belief normal distribution wouldn't great fit. reliability scores fall on bounded range, tend fall more towards 1 side or other of range. if tend high range, i'd predict lopped off @ end of range , have long tail low side, , vice versa if tend low range.


Comments

Popular posts from this blog

How to mention the localhost in android -

php - Calling a template part from a post -

c# - String.format() DateTime With Arabic culture -