Two-way or one-way MANOVA?

#1
Hi all!


I'm struggling with the following design:
I had 257 respondents filling in my survey with a manipulated website. The website was manipulated in 2 (source) X 2 (text) = 4 ways. So, in my data file i have 4 groups (for example, group 1: source A, text A, group 2: source B, text A).

Now i want to look if the score (depended variable) differs per group. Should i do a two way MANOVA with source and text as fixed factors? Or a one way manova with group as a fixed factor. My feelings says i should do the two-way, since I have two independed variables, but in the end i am curious to check my hypothesis: that group 1 scores better than group 3, and group 2 is better than group 4, which i can do with the one-way version. Can you help me?:)
 
#2
Hey,

1) firstly, u mean ANOVA not MANOVA.
2) it depends on your hypothesis, but I would preffer 2 fixed factors which models the two effects as separate "causes" and not splitting the sample into 4 (= lower N per group which may lead to lower power, I assume)

M
 
#3
Hi M,


Thank you for your answer. I'm sorry for the misunderstanding, but 'score' is meausured on 10 variables, that I computed to two averages. Therefore, i do need the MANOVA.

What i'm worrying about, I don't care if source A is better than source B, and i do not care if text A is better than text B. I want to prove, since it is my hypothesis, that the combination A-B is better than A-A and the combination B-A is better then B-B.


This is the output from the two-way MANOVA tells me that there are significant differences for text (typeFocus) and source (TypeSource) but that is not what i hope to find, i want to know what combination scores significantly higher, but maybe i can use the interaction effect to do so?



http://postimg.org/image/ewsfvkc1d/