Negative Binomial Regression - Is this scatterplot okay (pic)?

#1
Hi there,

One assumption of negative-binomial regression is linearity. I have this scatterplot of IV and DV?

Would you say that there is linearity? Is this scatterplot okay?
 

rogojel

TS Contributor
#2
hi,
maybe orthers will disagree, but, nope, to me this looks like you would need additional explanatory variables. that bulge in the middle looks like something that would ruin your fit whatever method you used.

regards
 

Dason

Ambassador to the humans
#3
hi,
maybe orthers will disagree, but, nope, to me this looks like you would need additional explanatory variables. that bulge in the middle looks like something that would ruin your fit whatever method you used.

regards
Possibly. But as always it's hard to tell in situations like this. Sometimes the seemingly "extra" variance that occurs in the middle of the bulge is just due to sample size issues. When you have a lot of data it's more common to see those "larger" values.

Code:
set.seed(100)

nlow <- 5
nmid <- 1000
nhigh <- 20
n <- sum(nlow, nmid, nhigh)

xlow <- rnorm(nlow, 2.5)
xmid <- rnorm(nmid, 7.5)
xhigh <- rnorm(nhigh, 12.5)

x <- c(xlow, xmid, xhigh)

# Equal variance
err <- rnorm(n, 0, 1)
y <- 3 + x*1/15 + err

plot(x,y)
View attachment 6553