Feature Request: Comparison of node predictability (r2) across samples

[gloriathan]

Hi, I'm wondering if it would be possible to add an extension to the network comparison functions in bggm to do nodewise comparison of node predictability between two networks/samples (e.g., male vs. female, depressed vs. non-depressed). Thank you!

[donaldRwilliams]

hi. thank you for opening the issue. It is much more convenient to provide code here. I did actually have a function in the package to compare R2, but removed it. That said, it can be done as follows

library(BGGM)
dat <- na.omit(BGGM::bfi)
g1 <- subset(dat, gender == 1)[,1:10]
g2 <- subset(dat, gender == 2)[,1:10]

fit1 <- estimate(g1) 
fit2 <- estimate(g2)

r21 <- predictability(fit1, iter = 500)
r22 <- predictability(fit2, iter = 500)

r2_diff <- 
sapply(1:ncol(g1), function(x){
  quantile(r22$scores[[x]] - r21$scores[[x]], c(0.025, 0.975))
} )

t(r2_diff)

That provides 95 % intervals for the predictability difference.

Ill add a function to do this in the future.

[gloriathan]

Thank you so much!

[donaldRwilliams]

No problem.