topic Re: How to Properly Specify a Mixed Model in Discussions

How to Properly Specify a Mixed Model

RMSJaguar755 — Tue, 04 Mar 2025 07:11:34 GMT

Hello,

I am currently analyzing EEG beta power data using a mixed model in JMP and would like to ensure that I am specifying the model correctly. My dataset is structured in long format, where each row represents a single observation with the following variables:

Subject	Group	Stimulus	Electrode	Time	Power
1	A	X	Cz	-0.4	1.23
1	A	X	Cz	-0.2	1.45
1	A	X	Pz	-0.4	0.56
...	...	...	...	...	...
1	A	Y	Cz	-0.4	2.91
...	...	...	...	...	...
2	B	X	Cz	-0.4	1.31
...	...	...	...	...	...
11	B	Y	Pz	-0.2	2.34

I would like to perform an analysis where:

Fixed effects include Group, Stimulus, and Electrode, along with their interactions.

Questions:

How should I correctly specify the repeated structure in JMP’s mixed model?

I would appreciate any guidance on how to properly structure this analysis in JMP. Thank you!

Re: How to Properly Specify a Mixed Model

Victor_G — Tue, 04 Mar 2025 10:49:19 GMT

Hi @RMSJaguar755,

Welcome in the Community !
Did you already had a look at JMP Help section related to repeated measures modeling : Example of Repeated Measures ?

It seems your data format is very close to the one from the example displayed in the JMP Help, where response is measured and repeated at different time points.
I'm not sure about the use of "Power" in your context, but it could be used as a covariate and/or random effect.

If you need more help/guidance, could you provide more details (and possibly an anonymized dataset to test some modeling approaches), so that other members of the Community can join this discussion and provide insights ?

Re: How to Properly Specify a Mixed Model

RMSJaguar755 — Tue, 04 Mar 2025 11:13:18 GMT

以下のような文面で返信するとよいでしょう。

Thank you for your rapid response! I have checked the JMP Help section on repeated measures modeling, and I believe my data structure is similar to the example provided there.

To clarify my analysis:

I have EEG beta-band power data of two groups (Responders, non-Responders).
Each subject was presented with 3 types of stimuli.
EEG was recorded from 32 electrodes.
Beta power was measured at 6 time points (-0.4s, -0.2s, 0s, 0.2s, 0.4s, 0.6s).

My goal is to perform a mixed model analysis treating Subject as a repeated measure, while analyzing the effects of Group (Responder/non-Responder), Stimulus (S1, S2, S3), and Electrode (Cz, Oz, P3, P4) on beta power over time. I am unsure about the best way to specify the repeated structure in the mixed model framework in JMP.

I have attached a simulated dataset in long format for reference.

How should I structure the mixed model appropriately in JMP? Should I treat Time as a repeated measure and include Subject as a random effect? I would appreciate any guidance on setting up the model effects correctly.

Thank you for your help!

Re: How to Properly Specify a Mixed Model

statman — Tue, 04 Mar 2025 15:23:03 GMT

To clarify a few things:

1. A mixed model is a condition where you are combining fixed and random effects.

2. In your case, your response is Power, your fixed effects are the effects associated with Group, Stimulus and Electrode and your random variable is Time? I'm not sure Time is a random effect here in terms of estimable and assignable in the model. What I question is whether Time is independent of the fixed effects. Do the fixed effects change between the different Times? If not, then Time is not independent of the fixed effects The fixed effects were constant when two different data points were gathered in time. In this case, you would have repeated measures. Repeated measures do not have DF's. They can be assessed (by looking at the within treatment variation), but you do not get DF's to model those effects. You can estimate enumerative statistics to describe that data (e.g., average, variance, slope) and evaluate whether the fixed effects affect those response variables. The repeated measures are not an appropriate estimate of the MSE.

Re: How to Properly Specify a Mixed Model

statman — Tue, 04 Mar 2025 15:26:05 GMT

Please post your data set in a JMP table.