Share
Explore BrainMass

T-Test, Effect Size, Power

For this exercise, let's assume that you are setting up this study, rather than being called in after the fact. A review of the clinic's records indicates that their 8-week diet produces an average weight loss of 3.1 kg, with a standard deviation of  2.4 kg. If we desire to measure a large effect (Cohen's d  0.8), then the mean difference between groups (our desired effect size) should be at least:

0.8 x 2.4 = 1.92 kg

QUESTIONS TO ANSWER
? Description of data analysis procedures (one paragraph)
? Selection/justification of your basic statistical approach
? Any modification made based on your test(s) of assumptions
? Description of findings (paragraph or two)
? Summary table comparing groups
? Accompanying narrative statement on CI of differences and t-test
? Description of effect size and power of the analysis
? Compute the Number of Subjects Needed to Achieve a Given Power and Effect

See the attached files.

Attachments

Solution Preview

The same text is in the attached file.

QUESTIONS TO ANSWER
? Description of data analysis procedures (one paragraph)
o Selection/justification of your basic statistical approach
o Test(s) of assumptions underlying the statistical approach
o Any modification made based on your test(s) of assumptions

I would do a 2-sample t-test without equal variances assumed (you could do it with equal variances assumed, it probably wouldn't make a difference because the standard deviations of the groups are pretty similar). The test statistic is calculated as:

So, in this situation, t = (-1.1111 - (-5.0111)) = 3.9/√2.567 = 7.697
...

$2.19