Reply to thread

Message

<blockquote data-quote="Bill Zebub" data-source="post: 8609661" data-attributes="member: 7031982">The fallacy of these &quot;sample size isn't large enough to make a difference&quot; arguments is that a small sample size only means it's a lot more swingy...not that that the effect isn't there.&nbsp; Yes, the small sample size means you are more likely to see no effect from the +1, but you are equally likely to see an even greater than expected result.For example, if 10 tables each have two fighters, one with 15 Strength and one with 16 Strength, and each table fights a million rounds of combat, they are all going to see roughly the statistically predicted difference.If each table only fights 10 rounds of combat, though, one or two tables will report no difference, one or two tables will report a huge difference, and the rest will report something in the middle.And, really, the more fundamental mistake here is that &quot;small sample size&quot; is only an issue if you are using the sample to generate your hypothesis.&nbsp; If I said, &quot;At last night's game we had two fighters, and the one with 16 Strength was WAY better than the one with 15 Strength&quot; then it would be justified to say that the sample size was too small.</blockquote>

[QUOTE="Bill Zebub, post: 8609661, member: 7031982"] The fallacy of these "sample size isn't large enough to make a difference" arguments is that a small sample size only means it's a lot more swingy...not that that the effect isn't there. Yes, the small sample size means you are more likely to see no effect from the +1, but you are equally likely to see an even greater than expected result. For example, if 10 tables each have two fighters, one with 15 Strength and one with 16 Strength, and each table fights a million rounds of combat, they are all going to see roughly the statistically predicted difference. If each table only fights 10 rounds of combat, though, one or two tables will report no difference, one or two tables will report a huge difference, and the rest will report something in the middle. And, really, the more fundamental mistake here is that "small sample size" is only an issue if you are using the sample to generate your hypothesis. If I said, "At last night's game we had two fighters, and the one with 16 Strength was WAY better than the one with 15 Strength" then it would be justified to say that the sample size was too small. [/QUOTE]

Verification