"Three-way ANOVA" on JMP?

[#1]

Hi all,

My two questions are rather straightforward. Perhaps the question already have been posted, if so you may share a link for that discussion.

1. Is it possible to conduct a three way interaction using JMP software? i.e. a so-called 3-way ANOVA, although it doesn't go by that name I think.
2. If you can conduct "3-way ANOVAs", which function(s) of JMP are you using?

My goal is to find out the significance values of the interactions of different plant traits that are coupled under treatment (disease) x accession (or genotype) x time. Time is a variable that have two defined time points only. If using only two different interaction parameters it can be conducted using the JMP Fit Model platform for making two-way ANOVA interaction. This I have done before.

If anyone is willing to give a helping hand, I will be most happy!

Best regards,

Jim

[#2 · Accepted solution]

Hi Jim,

Using Analyze > Fit Model you can specify your 3-way factorial ANOVA manually or using the Macro > Full Factorial after selecting your factors in the column list.

When you use the "Macro" for "Full Factorial" JMP will populate the Model Effects section with the main effects, 2-way interactions, and the 3-way interaction. Be sure that your predictor variables are all marked as Nominal.

I hope this helps!

Julian

[#3]

Hi Julian.

I am doing this same sort of 3 way interaction but with plants. My problem is that my 'Effect Summary' shows me no P-values, it is just blank. If I do a 2 way interaction it will show me P-values, so I do not understand what is going wrong when it turns into a 3 way interaction. My interaction factors, block, variety and year are all data type 'character' and modeling type 'nominal'. My degrees of freedom in 'Effect tests' seem to be on point. I would really appreciate your input, you make great YouTube videos so hopefully it is an easy fix for me.

Thank you

Laura

[#4]

Hi @Boots2011 ,

the reason you are not seeing any pvalues (or SD) is that there is only one observation for each unique combination of Block, Variety and Year, 

Notice that the R2 is exactly 1. Therefore, no variance left in the data that is not controlled for by the model.

Pending on the purpose and the context of the research question, perhaps worth modeling without the full factorial term (last three way interaction).

[#5]

Awesome, thank you so much for this explanation. I will move forward with just 2 interactions instead of 3. 

[#6 · Accepted solution]

Use Analyze > Fit Model, putting your response in Y box and then simultaneously highlight your three factors and select Full Factorial from Macros pull-down menu (located to the left of the Effects in Model section of the dialog box). This will fill your Effects in Model box with the three main-effects, the three two-way interactions and the three-way interaction between your factors.

[#7]

Hej!

Thanks a lot for the help! I solved the tasks some time ago.

Sorry for a late reply. I got carried away with the analyses and forgot to reply back.

Best regards,

Jim