ฝัง
- เผยแพร่เมื่อ 22 ก.ค. 2024
- I demonstrate how to perform an interaction contrast analysis in SPSS. This example is based on a 2x2 between-subjects ANOVA context. The main advantage of an interaction contrast is that you can specify the nature of the interaction on the basis of theory, which allows you, some would argue, to conduct a one-tailed interaction analysis - something that is not possible with a regular ANOVA.
I describe the nature of interactions in an accessible way in my textbook (Chapter 9): www.how2statsbook. Additionally, you should check out this video: • V9.7 - Simple main eff...
Is the pooled error term always the same for a 2 x 2 factorial ANOVA as for a 1 x 4 one-way ANOVA (which you use to compute the planned contrasts)?
You get a correspondence between the p-value for the 2 x 2 factorial interaction in the ANOVA, and the p-value for the contrast depicting the (constructed) interaction in the one-way.
But if you compute simple effects in a factorial ANOVA interaction using EMMEANS, will they correspond the means in a weighted contrast analysis in a one-way ANOVA (e.g., 1 -1 0 0)?
Hi
I have seen studies reporting df, F and SS as for an interaction based on planned contrasts. Normally the one-way ANOVA provides a "t" so does that happen? Please let me know?
Hi everyone, I am currently planning to conduct a mixed anova - within factor - 2 levels of equal importance, between factor - 3 levels of equal importance. How do I choose the contrasts for the between factor as I have no directed hypothesis?
Is it possible to a similar analysis with a dichotomous dependent variable? I'm trying to figure out how to essentially do a 2x2 ANOVA with contrast coding, but with a dichotomous DV, which isn't allowed with ANOVA.
Can contrasts be performed on repeated measures?
Yes. However, interaction contrasts with a within-subjects factor would be different to what I showed in this video. I'll see if I can figure it out.
Thank you so much!!
Thank you for making a tutorial video on (interaction) contrasts. Since the SPSS help is very limited, this should help a lot of people.
However, I think your statements are a bit dangerous and might lead to fraud. I think it isn't sufficiently clear from your explanation that the planned contrasts should really be planned. If people decide to do the contrast test because, like the video seems to suggest, the interaction p-value isn't significant, this would be fraudulent. The only good way to do this one-sided test is if the researcher has decided to do exactly this analysis before collecting the data.
I feel that saying things like "instead of being stuck with a nonsignificant effect" is a direct encouragement to commit scientific fraud. Sorry if this sounds harsh, but we should be responsible statisticians and educate not only on how to perform the methods, but also on the ethics of what we are doing.
He did say that you need theory in advance!!