I'm currently working on our thesis and I got stuck with the analysis part for I am not sure which statistical test I should use. My work had 5 groups. 1 positive and 1 negative control and 3 experimental groups. Is it right to use a 2-way ANOVA? And should the controls be included in the ANOVA? I'd like to compare if the experimental groups significantly varied from the control groups. What would be the best fitting test? Thank you so much.
$\begingroup$
$\endgroup$
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
4
From what you say so far it sounds like a one way ANOVA because you have mentioned only one independent variable: Group. This would have 5 levels. If you have other variables, that would add to the "ways".
You should definitely include the control group, otherwise you will not be able to compare them to the treatment. You can test your hypothesis with a contrast.
As for "best fitting test" that's impossible to say without looking at the data and the results of the models.
answered Jan 1, 2014 at 13:26
-
$\begingroup$ Thank you! :) But what if I included a pretest and a posttest? $\endgroup$ Commented Jan 2, 2014 at 10:10
-
1$\begingroup$ Then there are a number of choices: Gain scores, ANCOVA or possibly mixed models. $\endgroup$ Commented Jan 2, 2014 at 11:38
-
$\begingroup$ Why not a repeated measures ANOVA? $\endgroup$– BehacadCommented Jan 31, 2014 at 19:01
-
$\begingroup$ RM ANOVA makes assumptions that mixed (or multilevel) do not. See e.g Hedeker & Gibbons, Longitudinal Data Analysis $\endgroup$ Commented Jan 31, 2014 at 22:10