T-testing of two samples(question about t-statistic)

#1
Hi have two questions regarding an t-test assignment i did.

Basically the null hypothesis was that if sample x2 differed 35 off the mean from sample x1 was it reject-able?
The mean of sample x is 395
The mean of sample ý is 435.

So the issue is that i think the nominator for the t-statistic would go like this:
(395-435)-35.
Since x_mean - y_mean - (difference in hypothesized mean) would be the general form.
But my teacher writes it like 435-395-35, swapping x and y, and i don't understand why.

Second questions is that he does a two tailed CI(confidence interval), but for a test in difference were sample y is greater then sample x isn't it enough to only look at the higher bound?
 
Last edited:
#2
There is no presumption/knowledge that mean of x has to be more or less than mean of y, so 95% chance the "true" difference would locate within the (x_mean - y_mean) +- 2SD or (y_mean -x_mean) +- 2SD, that is why you can write it in both ways.

Same answer for the second question. If you have the knowledge one sample mean is more than another, then you can just using the higher bound.