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; } } }

Difference of R and JMP in Mixed Model

IkjuShin

Hi all,

I got some detailed question about Difference of R and JMP in Mixed Model.

1. She made test data using R (attached)

### Intention : Random Intercept SD 8

library(lme4)

set.seed(3)

group.inter = rnorm(5,sd=8)

group.slope = rnorm(5,sd=5)

xx = 4 + rnorm(100,sd=3)

yy = rep(group.inter,each=20) + xx * rep(group.slope,each=20) + rnorm(100,sd=3)

group1 = rep(c("A","B","C","D","E"),each=20)

summary(lmer(yy ~ xx+(xx|group1)))

2. Analysis Result is like below in R, the result is different

3. Random Effect(Intercept, Var) is too much different

-R : 56

-JMP : 674

1) I could not find why this difference happens

2) Which option should be adjustd or changed to maker two values equal ?

Thanks

Ikju

20190911-FitMixed.jmp

Accepted answers

jiancao

You'll get the same variance and covariance estimates if you turn off the Center Polynomials option (the default is on).

Notice that for these variance and covariance estimates JMP reports the standard errors while R shows the standard deviations.

I have also compared JMP's results to SAS Proc Mixed's. They match.

All comments

MikeD_Anderson

@IkjuShin - the confidence interval for the JMP Model contains the var priediction for the R value. I don't think you can say they're different. Note also that the p-value for the JMP model shows a lack of significance and the confidence interval for the term contains zero. I'd have the customer look at (or look yourself if your comfortable in R) the p-value and confidence interval for the R fit. It's possible they are trying to see significance in noise.

Someone, please feel free to correct me if my interpretation is incorrect here.

Best,

M

jiancao

You'll get the same variance and covariance estimates if you turn off the Center Polynomials option (the default is on).

Notice that for these variance and covariance estimates JMP reports the standard errors while R shows the standard deviations.

I have also compared JMP's results to SAS Proc Mixed's. They match.

jiancao

here is the data with my scrip saved.

20190911-FitMixed JC.jmp

IkjuShin

Hi Jiancao,
Thank you for your feedback.
We(I and Customer) got big present from you in Korean Thanksgiving day
Thanks again

jiancao

Have a good holiday! I wanted to pass this news along to you: it has been just decided that in JMP 15.1, for Mixed Model personality, Center Polynomials will be turned off by default.