Quote:
Originally Posted by Dyzalot
Exactly.
|
Seems like you've just contradicted yourself. If the data set is small, sample size within that data set is less likely to be.
Example: You have 6 early humans, and 1 interaction with 1 set of predator footprints. Later the predator attacks and 1 early human dies. That's not a small sample size. That's the
entire sample size available to them at that time.
What if you have 30 early humans, in 5 different packs of 6, with each pack having 1 interaction with 1 set of predator footprints, and later 1 predator attacks 1 of those packs? The sample size is now smaller, but still large in the context of the size of the overall data set.
What if you have 3000 not-quite-so early humans, in 500 different packs of 6, with each pack having 1 interaction with 1 set of predator footprints, and later 1 predator attacks 1 of those packs? Now you have a small sample size, at least in comparison to the earlier data sets.
My point: early humans weren't making decisions based on small sample sizes. They were making decisions based on large sample sizes (sample size being relative, of course, to the size of the overall data set).
To illustrate: 600 humans eat 1 type of fish. 1 human gets sick, 599 don't. Should all humans stop eating the fish? If we are only aware of the 1 person who got sick, and not the 599 who didn't, and make our decision accordingly, then we're making a decision based on small sample size (with or without confirmation bias or false perceptions). To reverse the scenario: 599 humans get sick eating this fish, 1 doesn't. If we are only aware of the person who didn't get sick before making a decision whether or not all humans should or shouldn't eat this fish, what is the likely decision we will make, with what consequences?
To relate to this thread: now, as at any time over the course of human history, humans make the best decisions based on large sample sizes. Determining if there is a problem with the OOTP game engine is no different. The data set can be as big as you want it to be, so collect as much data as possible, then see if there's a problem or not.
Quote:
Originally Posted by RchW
None of which has anything to do with confirmation bias and false perceptions. One can smell a lion if their life depends on it.  |
True (though we're getting into another data set here, aren't we? 6 early humans have 1 interaction with lion scent. Later lion attacks, killing 1 early human. Small or large sample size?
).
Real reason for the length of this reply: procrastination. Alas, back to work i go.