Friday, July 16, 2010

What kind of statistics test should I run on this data?

I have 7 participants who took three tests: a control test and two experimental tests. So all together I have 21 scores (each participant has 3 scores). Should I do an ANOVA? I need help please also explain why you would do this test to because I would like to actually understand.

What kind of statistics test should I run on this data?
Yes, you should use Analysis of Variance.





The null hypothesis for ANOVA is that all the means are equal to each other. The alternate hypothesis is that at least one of the means is different form the others. ANOVA takes into account of the multiple means and can be used for comparing any number of means.





If you were to use t-tests, for example, to test to see if there is a difference between the test means then you would have to perform three hypothesis tests, control vs treatment 1, control vs. treatment2 and treatment 1 vs treatment 2.





If you conduct each of theses tests at the 5% level then you have a 95% probability to draw the correct conclusion. With three independent tests you have (0.95) ^ 3 = 0.857375, meaning you have a 0.142625 probability of a Type I error.





Say you had four means to compare then you have a (0.95) ^ ( 4 choose 2) = 0.95 ^ 6 = 0.7350919 of a correct conclusion; 0.2649081 probability of a Type I error.





The ANOVA takes into account the multiple comparisons so the probability of a Type I error is the same as the significance level of the the test. If you do use multiple t -tests you can use multiple testing corrections such as Fisher's LSD (least significant difference) or Bonferroni's correction.


No comments:

Post a Comment