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; } } }

How to plot the estimates (fixed terms) of a linear mixed model

Catarina

Hello,

I have regular JMP (no Pro) and have used a linear mixed model using the REML option having as fixed terms 3 single and 3 crossed terms. All single terms used were nominal with 4, 3 and 2 levels respectively. I want to plot the estimates (fixed terms) as a standardized coefficient with respective associated 95%CI as significance level. However, I am unsure in the output where (or if) I can find this information (i.e. standardized estimate coeff of each fixed-term and associated 95%CI so I can plot them in a graph)? Under parameter estimates I get that information but for each fixed-term level, whereas I want the standardized coefficient per fixed-term and not per level within each term. I hope it makes sense!

Thank you!

Accepted answers

All comments

ron_horne

Hi @Catarina

did you try the Factor profiling menu under the red triangle as in the picture. it has a few options and from them there are extra options.

otherwise, all the different estimates of the model could be extracted from the effect screening menu as in the picture:

another way is to right click on the parameter estimates column go to Columns >> STD Beta

hope it works.

Catarina

Hi @ron_horne

Thank you so much for your suggestions. I can see you are using JMP Pro and unfortunately I don't have access to that version.

I have tried the Factor profiling but I am not sure I can extract standardized coeff from single and crossed fixed factors from there?

Factor Profiling.jpg

I also don't have access to Effect Screening

Effect screening.jpg

And in my version I don't have access to STD Beta under columns in the Parameter Estimates.

Columns STD Beta.jpg

Would you have any other suggestions?

ron_horne

Hi @Catarina ,

i am a bit surprised to hear these are pro features. i hope someone else in the forum can comment on that.

in any case, you should be able to standardize your variables prior to the analysis and run the regression. then the coefficients would be standardized.

Catarina

Thanks @ron_horne. The model I am using has nominal factors as fixed effects as well as crossed:

plant richness (1, 2, 3, 4);

Mic_dil (D0, D2, D6);

Drought (W, D);

plant richness x mic_dil;

drought x plant richness;

drought x mic_dil.

The model also has random effects.

If they would be continuous I would understand how to standardize them but I am not sure how to in this case? Thank you for your suggestion.

Mark_Bailey

I am confused. How do you standardize nominal or ordinal variables? Also, there is a parameter for each level, so what single estimate do you expect?

Catarina

I was hoping to get a parameter estimate for the fixed effect and not for the each level within each fixed effect. Something similar to a F statistic. But probably that does not make sense?

ron_horne

the only way this can make sense is if the effect is binary. then, there is just one coefficient....

I must admit that i have seen over the years people reducing multi categorical effects to binary for simplicity. in many cases without disclosing it properly and as part of bad practice overall.