Latin square with ordinal data

[#1]

Is it possible to analyze the following experiment in JMP? If so, how?

Experimental questions: 1) Did any treatments decrease plant disease severity or death compared to the untreated control? 2) Did any treatments delay disease onset compared to the untreated control?

-Tested 4 treatments on a plant pathogen against an untreated control

-Experimental design: 5x5 Latin square

-Assigned a disease severity score of 0, 1, 2, 3, or 4 (with 0 having no disease symptoms and 4 being dead; only integers were given)

-Disease severity was scored every 7 days throughout the experiment.

Unless I'm mistaken, rows and columns should be assigned as random variables for this analysis (to account for soil nutrient variation). Still, I need help finding a Personality that supports random variables when using ordinal scoring data. Also, can the treatments be compared to the untreated control in a Latin square design? Any recommendations?

[#2]

Hi @StatlearnR,

I'm not entirely sure about the design you have used (a datatable would have been easier to figure out), but you may find useful answers about similar questions here :

How to analyze a 4x4 Latin square design? 

Analyze Latin squares using JMP 

Repeated randomized 4 by 4 latin square 

It looks like you can analyze your data by specifying Treatment as a fixed effect in your model, and rows and columns of your LS design as random blocking effects, using the Fit Model platform. 

I did a simple dataset and demo trying to fit to your infos and context :

Datatable is attached with the script "Model" to launch the correct modeling, and "Fit Least Squares Yield" with the result of the modeling. Random Effects are however not supported by ordinal logistic models, so you might have to transform your response, or use another one. I added as an example a binning formula column of Yield with ordinal data, you can try launching ordinal logistic regression but it won't work with random blocking effects.

Hope this may help you. If not, please bring more information and if possible, an (anonymized) datatable.