Can I study 2- or 3-way models for a 4-way study?

#1
I've done a financial study that really has four independent variables and I'm trying to study the effects on trading performance. I've read that 4-way ANOVAs are rarely used because interactions are hard to understand. I could run six 2-way ANOVAs or four 3-way ANOVAs instead, but aren't I still at risk for having an unrecognized 4-way or 3-way interaction (even though I can barely comprehend what that would mean)?

Also, if I do run multiple 2- or 3-way ANOVAs, then would I have to correct for multiple comparisons as a result? Would the Holm method work for this?
 

Miner

TS Contributor
#2
The choice of the 4-way ANOVA has nothing to do with whether the 4-way interaction is understandable, but everything to do with the fact that you have 4 factors to analyze.

Run the 4-way ANOVA, and remove non-significant terms from the model. This increases the power for the remaining terms. Most high order interactions are very weak, may be spurious, and are rarely useful unless they occur in a chemical/nuclear environment.
 
#3
Touché. :)

As it turns out, I've found a plug-in online that allows me to do the 3-way ANOVA. I don't have software that computes a 4-way ANOVA. Should I be concerned?

Thanks!
 

Miner

TS Contributor
#4
I recommend sticking with the 4-way ANOVA. It is not that difficult to analyze manually by creating the formulas in Excel to create your own ANOVA table. I used to manually calculate them by hand.
 
#5
I recommend sticking with the 4-way ANOVA. It is not that difficult to analyze manually by creating the formulas in Excel to create your own ANOVA table. I used to manually calculate them by hand.
Can you provide a link that shows how to do this--or how to do 3 with guidelines to extrapolate to 4?

I haven't been able to find anything on how to do a 4-way ANOVA.