loveoforganic

-Account Deactivated-
10+ Year Member
Jan 30, 2009
4,222
13
Status
Hey guys,

For simple random sampling, literature I've read gave a sample size of 250 as a generally "safe" number to switch from matching, urn, etc. methods to SRS. I assume this is sample size per treatment condition (as opposed to total sample size), given that if # treatment conditions -> total sample size, each condition's sample would hardly be representative of the population, correct?

TYIA!
 
Last edited:

Ollie123

10+ Year Member
Feb 19, 2007
4,774
1,294
Status
Psychology Student
I looked into it a bit and couldn't find anything definitive. I would actually suspect it is 250 total n as long as you have a reasonable number of groups just because it is very very rare to see experimental research with > 200 per group but a whole lot of them seem to use SRS anyways.

Realistically, computers have pretty much minimized the effort needed to do things like urn randomization, so when in doubt I say go for it. Urns certainly shouldn't hurt anything. At the very least, picking your big ticket items and stratifying can achieve nearly the same thing. Just make sure you are pulling people OUT appropriately if it is a longitudinal study and you have attrition (lesson learned on that one....).
 
OP
L

loveoforganic

-Account Deactivated-
10+ Year Member
Jan 30, 2009
4,222
13
Status
Thanks :) 250/group seemed like an uncommonly large number, but I was getting my info from Project MATCH research, which is obviously on the rigorous end methodologically.