How to run a t-test for two groups with one group including only one value?
Example:
group A: 7, 8, 9
group B: 5
$\begingroup$
$\endgroup$
3
-
1$\begingroup$ This can be done only if you assume the two groups have the same variance. But the test would be very weak. $\endgroup$– Michael R. ChernickCommented Feb 27, 2018 at 2:47
-
$\begingroup$ @MichaelChernick With this assumption, it will be t.test(7:9, mu=5)? Thank you. $\endgroup$– Zhilong JiaCommented Feb 27, 2018 at 2:58
-
1$\begingroup$ $\mu$ is unknown and not necessarily 5. But 5 would be the estimate of the group B mean and 8 the estimate of the group A mean. $\endgroup$– Michael R. ChernickCommented Feb 27, 2018 at 5:28
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
No, your suggested code is quite wrong.
The fact that there is only one observation for group B is not a problem for the standard two-sample t-test. You just run it in R in the usual way:
A <- c(7,8,9)
B <- 5
t.test(A, B, var.equal=TRUE)
The option var.equal=TRUE tells t.test() to perform the traditional two-sample t-test with pooled variance. In this case, the population variance is estimated from group A alone since replicates for group B are not available. The t-test will be on 2 degrees of freedom.
answered Feb 27, 2018 at 4:00