Difference between revisions of "Talk:The Docks"
(adding drop data) |
|||
| Line 10: | Line 10: | ||
Are the big gaps in occurrence enough to warrant individual encounter percentages? It seems pretty clear that the Guard and Clerk are more common than the others, but should the smuggler also be singled out? I'll leave someone more mathematically minded to judge... --[[User:Jesus|Jesus]] 05:58, 14 August 2009 (UTC) | Are the big gaps in occurrence enough to warrant individual encounter percentages? It seems pretty clear that the Guard and Clerk are more common than the others, but should the smuggler also be singled out? I'll leave someone more mathematically minded to judge... --[[User:Jesus|Jesus]] 05:58, 14 August 2009 (UTC) | ||
| + | |||
| + | If you want to look at it as pure statistics, such differences with this sample size are not significant. What you are looking for is to determine if the difference between two dependent proportions is significant. For proportions Pa and Pb and sample size n the standard error is sqrt((Pa+Pb-(Pa-Pb)^2)/n). For 95% certainty we require a difference of at least twice that. If you do the math you'd see that the difference between 498 and 468 isn't significant, nor is the one between 548 and 498. In fact, even the difference between the 548 and 478 is barely significant. You need many more samples to make a 2% difference be significant.--[[User:Muhandes|Muhandes]] 10:13, 14 August 2009 (UTC) | ||
Latest revision as of 02:13, 14 August 2009
Data
big & tall shoreman (468) 76 big left hook / 98 docker's pants
old man from the sea (478) 103 bag of throwing starfish / 69 harbor pearl
pier security guard (547) 159 long arm / 130 SQUID: Tao of the Dow / 3 'Swift...' trilogy
small-time smuggler (498) 91 seventh of nine / 376 smuggler's box
stevedora (461) 67 unisex beanie / 55 work gloves / 151 jawbruiser
Shipping Clerk (548)
3000 turns
Are the big gaps in occurrence enough to warrant individual encounter percentages? It seems pretty clear that the Guard and Clerk are more common than the others, but should the smuggler also be singled out? I'll leave someone more mathematically minded to judge... --Jesus 05:58, 14 August 2009 (UTC)
If you want to look at it as pure statistics, such differences with this sample size are not significant. What you are looking for is to determine if the difference between two dependent proportions is significant. For proportions Pa and Pb and sample size n the standard error is sqrt((Pa+Pb-(Pa-Pb)^2)/n). For 95% certainty we require a difference of at least twice that. If you do the math you'd see that the difference between 498 and 468 isn't significant, nor is the one between 548 and 498. In fact, even the difference between the 548 and 478 is barely significant. You need many more samples to make a 2% difference be significant.--Muhandes 10:13, 14 August 2009 (UTC)
