Scaled estimates plot after REML analysis

[#1]

Hello, 

For a DOE analysis, I am using the REML method because I have a random effect. However, using this method, JMP does not generate a scaled estimates plot in the report, nor a Pareto plot. This option is not available for REML, even though a table with the parameter estimates is provided. Any suggestion on how I can generate a plot similar to scaled estimates plot after REML analysis? 

Thank you 

Sara 

[#2]

Hi @SaraA ,

Answered this on your other post but putting the answer here in case someone else comes across the same question:

Because of the nature of the method, REML focuses on variance components and (unlike methods like Least Squares) does not perform standardisation (which is a pre-processing step in the workflow for modelling) - there's a lot of additional complexity with mixed models and trying to standardize them (do you standardise globally or within groups? how do you standardize a variable that is different between two groups?).

Thanks,

Ben

[#3]

But the REML analysis does provide a table with parameter estimates, so why does it not provide something similar to this that you get in the standard least squares analysis? 

[#4]

Hi @SaraA ,

Those are the coefficients of the model parameters that every model can make - the difference is that to make the scaled estimates that has to be done beforehand by standardizing the parameters - in a simple linear model like SLS that's easy, in REML it's complex and not appropriate to apply, you can't scale estimates from a model if the standardization hasn't been done beforehand, and REML doesn't have that as part of it's workflow (for good reason).

Thanks,

Ben

[#5]

Hi @Ben_BarrIngh 

Is there a similar plot I could make using the parameter estimates (not scaled estimates) from the REML? I think it is a good plot for visualization of the data. Or does it not make sense to do that for REML?  

Thank you 

Sara

[#6]

Hi @SaraA ,

You could always right click the estimates table in the model report and 'Make into Data Table' then use Graph Builder to fit the estimate vs. the fixed or repeated effects - this would potentially make a very large visual so I don't know if it would be too complex.

Thanks,

Ben

[#7]

Thank you @Ben_BarrIngh, this works. 

[#8]

Hi @SaraA, 

No problem!

Thanks,

Ben

[#9]

Hi @SaraA  @Ben_BarrIngh :  In my experience designing experiments in JMP (via the DOE menu, with or without random effects) the parameters for the fixed continuous factors by default are based on the the scaled factors. You can verify if this is true, or not, via Estimates/Show Prediction Expression in the red triangle pull-down menu.

[#10]

Hi @MRB3855 

In the JMP DOE analysis using standard least squares, the parameter estimates are indeed identical to the scaled estimates as far as I can see. For the analysis with random effects/REML, the estimates are indeed the same as in the prediction expression. But than I don't understand why JMP does not generate this graph for REML analysis, the same way it does for Standard Least squares.... I have found a way to make a similar graph manually as suggested by @Ben_BarrIngh, but it is not ideal