You're looking at Less Wrong's discussion board. This includes all posts, including those that haven't been promoted to the front page yet. For more information, see About Less Wrong.

gwern comments on Open Thread April 8 - April 14 2014 - Less Wrong Discussion

3 Post author: Tenoke 08 April 2014 11:11AM

You are viewing a comment permalink. View the original post to see all comments and the full post content.

Article Navigation

Comments (242)
Tags: open_thread

Comments (242)

You are viewing a single comment's thread. Show more comments above.

Comment author: gwern 13 April 2014 03:01:23AM *  1 point [-]

Have LWers ever used Usenet? By that, I mean: connected to a NNTP server (not Google Groups) with a newsreader to read discussions and perhaps comment (not solely download movies & files).

Your age is:

I am curious about the age-distribution of Usenet use: I get the feeling that there is a very sharp fall in Usenet age such that all nerds who grew up in the '70s-'80s used Usenet, but nerd teens in the mid-'90s to now have zero usage of it except for a rare few who know it as a better BitTorrent.

Submitting...

Comment author: gwern 27 February 2015 02:30:35AM *  0 points [-]

A belated analysis: hypothesis confirmed to my satisfaction - as expected, age is strongly related to Usenet-familiarity. (There's even a hint of a quick shift in the histograms.)

# <https://www.dropbox.com/s/d5kddi1ckl3qbxk/2015-02-26-lw-usenet.csv>
usenet <- read.csv("2015-02-26-lw-usenet.csv", header=FALSE)
usenet2 <- data.frame(Usenet=usenet[1:67,]$V3, Age=usenet[68:134,]$V3)
## the default LW poll encoding is yes=0, no=1; this is very confusing, so let's reverse it
usenet2$Usenet <- (usenet2$Usenet==0)
wilcox.test(Age ~ Usenet, data=usenet2)
#
# Wilcoxon rank sum test with continuity correction
#
# data: Age by Usenet
# W = 183.5, p-value = 3.893e-06
g <- glm(Usenet ~ Age, data=usenet2, family="binomial"); summary(g)
# ...Coefficients:
# Estimate Std. Error z value Pr(>|z|)
# (Intercept) -7.33329350 1.93381008 -3.79215 0.00014935
# Age 0.24311715 0.06695693 3.63095 0.00028238
## alternative plot:
# with(usenet2, plot(Age,Usenet,xlab="Age",ylab="Probability of Usenet familiarity"))
# curve(predict(g,data.frame(Age=x),type="resp"),add=TRUE)
# points(usenet2$Age,fitted(g),pch=20)
library(popbio)
with(usenet2, logi.hist.plot(Age,Usenet,boxp=FALSE,type="hist",col="gray"))
## <https://i.imgur.com/W6DT9Tu.png>

Histograms of distribution of age split by Usenet-familiarity, with overlaid logistic regression line

## specific example: 51yo vs 20yo probabilities based on the model: predict(g, data.frame(Age=51), type="response")
# 1
# 0.9937299491
predict(g, data.frame(Age=20), type="response")
# 1
# 0.07791991187

(If anyone is curious, my original motive was wondering about Satoshi Nakamoto & Nick Szabo - both are familiar with and have used Usenet. We already know Szabo is old and very similar to LWers, so being Usenet-familiar turns out to be entirely ordinary as I guessed, but if Satoshi Nakamoto were a young college student as some people thought, then being Usenet-familiar is pretty surprising.)

Powered by Reddit Powered by Reddit