0
$\begingroup$

I have read that paired $t$-test works assuming a null hypothesis with equality, i.e: $$H_0 : \mu = \mu_0 $$ Is there any way to do the same test with an inequality, such as: $$H_0 : \mu \geq \mu_0 $$ Or there is a (~another) statistical test to do the comparison? Thanks.

$\endgroup$
  • $\begingroup$ You may want to look into equivalence testing which, essentially, reverses the classical roles of the null and alternative hypotheses. $\endgroup$ – if_the_correlations_are_zero Jan 11 at 15:48
0
$\begingroup$

In R with t.test using data mtcars as toy example

> t.test(x=mtcars$mpg,
       y=mtcars$mpg+rnorm(32,3,2),
       alternative="less",
       paired=T)

    Paired t-test

data:  mtcars$mpg and mtcars$mpg + rnorm(32, 3, 2)
t = -9.6243, df = 31, p-value = 3.967e-11
alternative hypothesis: true difference in means is less than 0
95 percent confidence interval:
      -Inf -2.468995
sample estimates:
mean of the differences 
              -2.996977
$\endgroup$
0
$\begingroup$

'Paired' does not imply a specific null hypothesis, it just means that the two measures are derived from the same sample instances.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.