I’m thigh deep in social network analysis here, and working up a grand post that summarized the whole scope, vision, and conclusion of #SciFund in terms of science engagement. It’s going to be hot, heavy, and luscious. Seriously.
In the meantime, I thought I’d seed another little statistical tidbitlet that looks intriguing. This is actually an analysis that I’ve dropped for reasons that will become clear in a later post, as I think it’s actually hiding some things, but, it’s a fun one that I just can’t let go by.
Question – how much are your friends worth?
Answer: 2 DOLLARS!!!!
How do I get this? A quick generalized linear model (quasipoisson error, linear identity link function) looking at total raised as predicted by number of Twitter followers and Facebook friends. Both end up being different from 0 (LR Chisq Tests, LR Chisq = 7.984 and 4.922 respectively, with p=0.047 and p=0.027 respectively). And here’s the kicker:
Call: glm(formula = total ~ TwitterFollowers + FacebookFriends, family = quasipoisson(link = "identity"), data = projects) Deviance Residuals: Min 1Q Median 3Q Max -69.13 -19.13 -2.69 12.71 97.38 Coefficients: Estimate Std. Error t value Pr(>|t|) (Intercept) 426.3531 383.4634 1.112 0.2757 TwitterFollowers 2.0229 0.8286 2.441 0.0212 * FacebookFriends 2.4087 1.0431 2.309 0.0285 * --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Now, this doesn’t fit the data super-duper well (R2 of observed v. predicted = 0.24 – blech), but, it is very intriguing. I’ll discuss next time how I clean this up (i.e., Facebook friends is not a good measure of scientific fanbase, and we need to be looking at pageviews, not total) – but, in the meantime, go look at your friends or followers list, and see the sea of glistening $2 bills flapping in the wind.
Oh, and, if you haven’t signed up for this round, sign-up closes on March 31st for #SciFund Round 2!!!!!!