Actuarial Outpost random sampling
 Register Blogs Wiki FAQ Calendar Search Today's Posts Mark Forums Read
 FlashChat Actuarial Discussion Preliminary Exams CAS/SOA Exams Cyberchat Around the World Suggestions

 Upload your resume securely at https://www.dwsimpson.com to be contacted when new jobs meet your skills and objectives.

 General Actuarial Non-Specific Actuarial Topics - Before posting a thread, please browse over our other sections to see if there is a better fit, such as Careers - Employment, Actuarial Science Universities Forum or any of our other 100+ forums.

#1
01-30-2018, 01:25 PM
 Frustrated & Unmotivated CAS SOA Non-Actuary Join Date: Jun 2015 Posts: 14
random sampling

what are the advantages (if any) of using random samples vs the entire population if the computing power exists to analyze the whole population?

e.g. you want to compare an insured of Class X, Subclass Y, territory Z with other such XYZ'ers?

Assume also the randomizer is sufficiently random to avoid selection bias.

Background:
We have a tool that does this comparison in-house using random sampling. it was created years ago. I'm wondering why we shouldn't just compare the whole population. studying this was years ago, so I forget all this. Yet i still have a needle in my brain that is telling me that Random Sampling is better.
#2
01-30-2018, 01:31 PM
 Frustrated & Unmotivated CAS SOA Non-Actuary Join Date: Jun 2015 Posts: 14

Moved from P&C for greater exposure. Figure it's a general stats question?
#3
01-30-2018, 02:00 PM
 Guinness Member SOA Join Date: Jul 2016 Posts: 43

Barring computational efficiency, one thing that springs to mind is using such a partition to keep yourself from over-interpreting random variation. You could use some random sample for ideation and another to test whether the idea really makes sense. Another thing you could always do some bootstrapped statistics to check variablity of various statistics: mean, standard deviation, etc...
#4
02-13-2018, 09:41 PM
 PAC Member CAS AAA Join Date: Jan 2003 Posts: 246

Over fitting... general modeling concept, should have a strategy to stratify (with random sampling) train/test/validate data sets.