Hello everyone,
I have a little problem regarding interpreting my results of pairwise comparisons (one-way ANOVA) for my masters thesis.
Situations is as follows:
I have the following three variables:
1. perceived social status (PSS, dependent variable)
2. ECO (independent and binary variable: ecological good (ECO_02) vs. non ecological good (ECO_01))
3. KS (collective guilt; moderator variable, also binary: low collective guilt (KS_M = 0) vs high collective guilt (KS_M = 1))
I have the following two hypothesis:
1. The consumption of ecological good increases the PSS of a person more than the consumption of non ecological good.
2. The influence of an ecological good on the PSS is amplified by collective guilt.
I want to test both hypothesis.
With which table (seen in both pictures) is a testing of the hypothesis better? Or is a reformulation of the hypothesis needed?
Thank you very much!
I think that ANOVA is not the suitable tool to treat your data of binary type. You need to use nonparametric tools. I think that logistic regression is suitable to analyze your data.
Hello Ali,
The second table ("Version_2", though the table images were _very_ small, and difficult to discern!) shows:
1. Eco2 > Eco1 as a direct influence ("main effect") for score on your DV, PSS (p = .0002). That supports your first hypothesis.
2. This difference (Eco2 vs. Eco1) is significant only when the proposed moderator, KS_M, is "high" (= 1) (p < .0001). The Eco2 vs. Eco1 difference is not significant when KS_M is "low" (= 0) (p = .2748). That supports both your second hypothesis and affirms that there is an interaction of Eco and KS.
Good luck with your work.
David Morse , "and affirms that there is an interaction of Eco and KS.": no, it does not. If you want to know if the data support an interaction (difference in Eco-effects between levels of KS), then it should be tested.
There is an important issue with the analysis and of interactions, particularly regarding their interpretation (e.g.if there is functional synergism or antagonism). Here it makes a huge difference it the effects are considered additive or multiplicative (and what the true relation of the effects is).
Jochen Wilhelm ,
The general principle applies in an anova setting: when interaction is present, differences in levels of one factor are inconsistent across levels of at least one other factor.
The fact that the two simple effects tests are not identical (in difference) demonstrates that SS(interaction) is guaranteed to be non-zero...that "affirms" the interaction, at least in the sample. Note I did not say this was a formal test of interaction, nor did I say it provided a p-value for the test of interaction.
Hope all are staying well.
David Morse , thank you for your response. We certainly agree here. It really was not clear to me when you meant sample statistics and when expectations or population parameters.
There's still a potential problem with the DV which I assume is PSS if The DV is not continuous then none of this makes any sense because ols. methods are not appropriate. So someone that knows please tell us about the DV WHAT KIND OF VARIABLE IS IT?. Best wishes David Booth
- Badges
- Science topic
- Similar topics
- Chemistry
- Pharmacology
- Pharmaceuticals
- Drugs