[R] Please Help pairwise.t.test!!

[Peter Flom]

Chunhao Tu <tu_chunhao at yahoo.com> wrote
>Hi R users,
>My question is, If I have 3 groups, A, B, C and I know mean of A =20, B=21,
>and C=20.5 and I also know the 
>standard error of A =1.1, B=2.2, C=3.2. Plus, I know A has 30 observations,
>B has 78, C has 45. But I do not have the raw data.
> 
>Can I use pairwise.t.test to conduct a Bonferroni test? If yes, Could you
>give me a hint? 

If the data in each group is normally distributed, you could use the mean
and sd and N in rnorm and then test the results.  But if you don't know that,
then I think it's a bit up in the air.

A - N = 30, sem = 1.1, mean = 20   sd = 1.1*30^.5 = 6.02
B - N = 78, sem = 2.2, mean = 21   sd = 2.2*78^.5 = 19.43
C - N = 45, sem = 3.2, mean = 20.5, sd = 3.2*45^.5 = 21.46

let's do just A and B, to keep it a bit simpler:

if both are normal:
A <- rnorm(n = 30, mean = 20, sd = 6.02)
B <- rnorm(n = 78, mean = 21, sd = 19.43)
t.test(A,B)

t = .927

but if A is uniform and B is normal

A <- rnorm(n = 30, mean = 20, sd = 6.02)
B <- runif(n = 78, mean = 21, sd = 19.43)
t.test(A,B)

t = 0.42

HTH, and hope I didn't mess up one of the calculations due to insufficient caffeine intake

Peter


Peter L. Flom, PhD
Statistical Consultant
www DOT peterflomconsulting DOT com