`sample()` gives different values with same `set.seed()`

Question

I was creating some random samples and plotting them and noticed a strange behavior. Sampled values were different after loading ggplot2:

set.seed(111)
library(ggplot2)
sample(1:10, 10)
# [1]  8  4  5  3  7  1  6  2 10  9

set.seed(111)
sample(1:10, 10)
#  [1]  6  7  3  4  8 10  1  2  9  5

I can avoid this behavior easily enough, but is there any reason for ggplot2 to change the seed value?

Norbert

It isn't about changing the seed value - it's about getting a random number (before you). – Hiller

Note that if you'd tested the repeatability of both results, it would have been a strong hint that ggplot2 was doing something (repeatable) with the random seed. Experiments are always good :-) – Subtitle

Answer 1

I think I saw some discussion of this in one of the R chat rooms: ggplot2 calls the random number generator in order to decide whether/which tip it wants to offer.

In particular, this is ggplot2:::.onAttach:

function (...) 
{
    if (!interactive() || stats::runif(1) > 0.1) 
        return()
    tips <- c("Need help? Try the ggplot2 mailing list: http://groups.google.com/group/ggplot2.", 
        paste("Find out what's changed in ggplot2 with\n", "news(Version == \"", 
            utils::packageVersion("ggplot2"), "\", package = \"ggplot2\")", 
            sep = ""), "Use suppressPackageStartupMessages to eliminate package startup messages.")
    tip <- sample(tips, 1)
    packageStartupMessage(tip)
}

It's sort of amusing that one of the randomly generated tips tells you how to turn off the tips ...

Cochineal

I'v never experienced ggplot2 giving any tips at startup or on attach, but that makes sense if it has that option. Thanks for the clarification. – Norbert

Well, you only get a tip 10% of the time ... and only in recent versions, I think. – Cochineal

Shouldn't it tip at least 15%? – Sirajuddaula