Reputation: 13
I have several data summaries for paired t-test from three difference sources (hospitals), which means I don't have the whole sample data. What I've got to perform the paired t-test is the mean (difference of each paired sample data).
The data looks like the table below (each row is one patient) and I have the mean(diff)
and sd(diff)
.
Is there any function with which I could easily perform paired t-test and get the 95% CI?
There is just no way of getting the whole data set, due to patient privacy concerns.
3 months 6 months diff
1 3 -2
2 1 1
5 9 -4
Upvotes: 1
Views: 578
Reputation: 226871
A paired t-test is exactly equivalent to a t-test that tests the mean of the differences against a null hypothesis of zero.
diffs <- c(-2,1,-4)
m <- mean(diffs)
s <- sd(diffs)
n <- length(diffs)
Now suppose you don't know diffs
, just m
, s
and n
(all three are necessary)
pval <- 2*pt(abs(m/s),lower.tail=FALSE,df=n)
## 0.555
qq <- qt(0.975,df=n)
ci <- m + c(-1,1)*qq*s
## [1] -9.675648 6.342314
Upvotes: 2