How big is "sufficient" sample size? I am currently doing some tests with some specific parameter settings to characterize a process.
I roughly remember that a widely used rule of thumb for lab tests and whatnot is a sample size of 5 or 6. Even my tutor said to do the same test 5 times. It's what were told in two of our undergraduate courses. 
The thing is, now that I am conducting some experiments on my own, I'd like to have a citation on that. Although I like stats, my thesis is in the field of mechanical engineering and thus this is all for a nice detail. I don't need it, but would really like to try and write down anyway. 
Does anyone have a legitimate source for that? Or could someone give me a few pointers on how to give a short conclusive explanation for that minimum sample size?
 A: There is no prior justification for claim: n = 5 is sufficient sample size. Such rules of thumb should be used with caution. It always depends on situation. 
Sometimes such size is sufficient, sometimes clearly not. In some fields of study (mainly in ecology) are researches spending lot of time by data collection. They think that "bigger n" equals "better results". You should focus, whether your values reflect the variability propagated by studied phenomena. If studied phenomenon has small variability in measured values, usually you need to collect smaller sample size (and vice versa).
Imagine the situation:
You have collected 1000 values for your research. The obtianing of such data is time demanding (and also not cheap). Think about the changing of variability
in your data as sample size increases. At the beginng, each added value will cause big differences in variability,but as process continues, the adding of new values will cause only slight changing in variability for whole dataset.
The sufficient sample size lies somewhere in the point when the differences in variability values are non-striking.
# full dataset
x <- rnorm(1000)

# computation of variability after adding each value
my.var <- numeric(1000)
for (i in 1:length(x))  {
my.var[i] <- var(x[1:i])
}

# to see what is happening...
plot(my.var, type = "l",
xlab = "sample size", ylab="sample variability", col = "red", lwd = 2)


It this case, the "sufficient" sample size seems to be 150-170 values (which is substantially less than 1000.
