0
$\begingroup$

I am designing a study for my project. I wanted to test if music affects reading comprehension. This study will be a between group design. The independent variable is type of music and the dependent variable is the score on a reading comprehension task.

Half of participants will be randomly assigned to a room with classical (soft) music, and the remaining half of participants will be assigned to the room with rock (hard) music.

Would this be a t-test? Then what kind of t-test should I need to use?

Or is this a one way ANOVA?

$\endgroup$
1
  • $\begingroup$ Then how could i make simple design? I am still studying stat... Could anyone explain why do i need to use unpaired t test? and what will happen if i don't? $\endgroup$ – user44491 Apr 26 '14 at 15:41
1
$\begingroup$

Your design involves two independent samples, so you will be conducting an unpaired test. An unpaired t test would be appropriate to infer mean difference. You could also use a rank sum test, and if the two groups have similar (univariate) distributions of scores, you could infer median difference.

$\endgroup$
7
  • $\begingroup$ Beat me by a couple of seconds, but here's my contribution anyway: Think of ANOVA as a larger-scale t-test, because you'll be simultaneously comparing multiple means (usually more than two). $\endgroup$ – Waldir Leoncio Apr 25 '14 at 20:52
  • 1
    $\begingroup$ Albeit ANOVA comes with stricter assumptions than a t test. ANOVA assumes (1) near normality of the outcome variable (scores in this case), and (2) equal variances of the outcome in both groups. $\endgroup$ – Alexis Apr 25 '14 at 20:57
  • $\begingroup$ Right. Parametric tests can be quite annoying. :P $\endgroup$ – Waldir Leoncio Apr 25 '14 at 21:01
  • $\begingroup$ Although the t test (paired or unpaired) gets away with not assuming normality/near-normality of the outcome variable (or difference in outcome variables, for paired test) by assuming asymptotic normality of the sample mean. $\endgroup$ – Alexis Apr 25 '14 at 21:04
  • $\begingroup$ Then how could i make simple design? I am still studying stat... Could anyone explain why do i need to use unpaired t test? and what will happen if i don't? $\endgroup$ – user44491 Apr 26 '14 at 15:43
1
$\begingroup$

Sounds like you want something like an independent-samples t-test, but which exactly is most appropriate depends on the details of your data. Student's t-test involves several assumptions that may not hold in your data. It's fairly robust against some violations, but sensitive to others. The distributions of your groups' data will be important to consider, and probably sample size as well.

Variations on this (parametric) t-test include the , bootstrap tests, permutation tests, and probably very many others. A one-way parametric ANOVA with only your two groups ought to give you basically equivalent results to those of Student's t-test, but it's generally expected to report results of a t-test in your case.

$\endgroup$

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.