## LESSWRONGLW

Open thread, August 19-25, 2013

I have made it up to episode 5 of Umineko, and I've found one incident in particular unusually easy to resolve (easy enough that though the answer hasn't been suggested by anyone in-game, I am sure that I know how it was/could be done); I'm wondering how much it is due to specialized knowledge and whether it really looks harder to other people. (Because of the curse of knowledge, it's now difficult for me to see whether the puzzle really is as trivial as it looks to me.) So, a little poll, even though LWers are not the best people to ask.

In episode 5, a... (read more)

Showing 3 of 15 replies (Click to show all)

Creuncf gur fyvc bs cncre ybbxrq fbzrguvat yvxr guvf. (Qrfvtavat na nzovtenz jbhyq or nanybtbhf gb svaqvat zhygvcyr zrffntrf jvgu gur fnzr unfu.)

0gwern6yThanks for all the poll submissions. I decided since I just finished Umineko, this is a good time to analyze the 49 responses. The gist is that the direction seems to be as predicted and the effect size reasonable (odds-ratio of 1.77), but not big enough to yield any impressive level of statistical-significance (p=0.24): R> poll <- read.csv("http://dl.dropboxusercontent.com/u/182368464/umineko-poll.csv") R> library(ordinal) R> summary(clm(as.ordered(Certainty) ~ Crypto, data=poll)) formula: as.ordered(Certainty) ~ Crypto data: poll link threshold nobs logLik AIC niter max.grad cond.H logit flexible 48 -30.58 67.16 5(0) 5.28e-09 2.9e+01 Coefficients: Estimate Std. Error z value Pr(>|z|) Crypto 0.571 0.491 1.16 0.24 Threshold coefficients: Estimate Std. Error z value 0|1 1.988 0.708 2.81 1|2 3.075 0.822 3.74 (1 observation deleted due to missingness) R> exp(0.571) [1] 1.77 Or if you prefer, a linear regression: R> summary(lm(Certainty ~ Crypto, data=poll)) Call: lm(formula = Certainty ~ Crypto, data = poll) Residuals: Min 1Q Median 3Q Max -0.409 -0.287 -0.287 -0.164 1.836 Coefficients: Estimate Std. Error t value Pr(>|t|) (Intercept) 0.164 0.151 1.09 0.28 Crypto 0.122 0.117 1.05 0.30
1gjm6yI'm rather alarmed at how many people appear to have said they're very sure they know how he did it, on (I assume, but I think it's pretty clear) the basis of having thought of one very credible way he could have done it. I'm going to be optimistic and suppose that all those people thought something like "Although gwern asked how sure we are that we know how it was done, context suggests that the puzzle is really 'find a way to do it' rather than 'identify the specific way used in this case', so I'll say 'very' even though for all I know there could be other ways'. (For what it's worth, I pedantically chose the "middle" option for that question, but I found the same obvious solution as everyone else.)

# Open thread, August 19-25, 2013

1 min read19th Aug 2013326 comments

# 4

If it's worth saying, but not worth its own post (even in Discussion), then it goes here.