Hi All, I've had 2 questions from a customer trying to do some work in JMP. These are: * Can a Wilcoxon / Kruskal Wallace test be performed on Ordinal data directly in JMP? I appreciate it can be treated as continuous dataset to achieve the same result just wondered if it can be done without this step! * When generating a mixed model and using REML, is it possible to impose a correlation structure i.e. autoregressive on it? Thanks in advance Stuart

nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

Wilcoxon and REML Questions

Stuart_Little

Hi All,

I've had 2 questions from a customer trying to do some work in JMP. These are:

Can a Wilcoxon / Kruskal Wallace test be performed on Ordinal data directly in JMP? I appreciate it can be treated as continuous dataset to achieve the same result just wondered if it can be done without this step!
When generating a mixed model and using REML, is it possible to impose a correlation structure i.e. autoregressive on it?

Thanks in advance

Stuart

Accepted answers

jiancao

For #2, yes, there is an AR(1) -autoregressive of order 1- for within-subject covariances. It assumes a correlation between two adjacent measurements with geometric progression.

https://www.jmp.com/support/help/en/16.2/#page/jmp/repeated-covariance-structures.shtml#ww1273585

It is one of my favorites since it can be used together with an additional random effect to account for between-subject variability.

All comments

Phil_Kay

Hi @Stuart_Little ,

Not sure about 1.

As for 2. I think you would need the Mixed Model personality in JMP Pro for specifying correlation structures. @jiancao is the expert on this stuff.

Hope that helps in some way?!

Phil

jiancao

For #2, yes, there is an AR(1) -autoregressive of order 1- for within-subject covariances. It assumes a correlation between two adjacent measurements with geometric progression.

https://www.jmp.com/support/help/en/16.2/#page/jmp/repeated-covariance-structures.shtml#ww1273585

It is one of my favorites since it can be used together with an additional random effect to account for between-subject variability.

karencopeland

For #1 in oneway (ordinal X and continuous Y) the Wilcoxon /KW test is available...is that what you are looking for?

Stuart_Little

Thanks @jiancao, that answers #2 perfectly.

Stuart_Little

Thanks @karencopeland for your response. Sadly, it's not the situation the customer has - their data is Ordinal Y and Categorical X. I suspect that this one may not be possible, sadly.

karencopeland

Look up Cochran-Armitage trend test...my quick look at Agresti suggests that might be the appropriate test (but review and make sure that is the case) and that is in the contingency platform.