statistics for behavioral sciences
posted by troy on .
In a one- to two-page Microsoft Word document, for each hypothesis listed indicate:
Whether the appropriate analysis would be a one-tailed test or a two-tailed test.
A type I and type II error, given the context of the hypothesis.
The three treatments tested differ in how well they work.
Children receiving supportive therapy will get worse over time (at posttreatment).
Girls will do better than boys with treatment (any treatment).
To determine whether or not to use a two-tailed test versus a one-tailed test depends on the alternate hypothesis. If the alternate hypothesis shows direction (less than or greater than), a one-tailed test is used. If the alternate hypothesis shows no specific direction, then a two-tailed test is used. Usually a problem will give you a clue as to whether to use a one-tailed or two-tailed test. For example, the problem might ask only if there is a difference (two-tailed), or a problem might look for less than a population value or greater than a population value (one-tailed).
You can reduce Type I errors (rejecting the null when it's true) by setting the alpha at a lower level; however, when you do that, you increase the probability of making a Type II error (accepting the null and it's false). Most researchers try to minimize errors by setting alpha to .05.
I hope these few explanations will help.