f



MIxed effectds model

Hello,

I am conducting the linear mixed effects analysis by SPSS. I have few questions and highly appreciate your help:

1- How can I compare two linear mixed effects models to find the best model?
I've been told to use the t-test for coefficients but how? What is the SPSS command?

I need to compare one simple mixed effects model with the Univariate one.

2- How can I obtain the fitted values for my model?

3- How can I plot my mixed effects model in a way that it also shows confidence interval. I plotted the predicted value versus my independent variable but it looks like the linear line and it does not show me the confidence interval.

Thanks so much for your help
0
Mojy
8/23/2016 7:43:24 AM
comp.soft-sys.stat.spss 5679 articles. 0 followers. Post Follow

2 Replies
200 Views

Similar Articles

[PageSpeed] 3

On Tue, 23 Aug 2016 00:43:24 -0700 (PDT), Mojy <mgmoghaddam@gmail.com>
wrote:

>Hello,
>
>I am conducting the linear mixed effects analysis by SPSS. I have few questions and highly appreciate your help:
>
>1- How can I compare two linear mixed effects models to find the best model?
>I've been told to use the t-test for coefficients but how? What is the SPSS command?
>
>I need to compare one simple mixed effects model with the Univariate one.

For a nested comparison, you have an F test from the 
change in mean-square and change in d.f.  If you are using
regression, that is among the options -- after entering a Step. 

For non-nested comparisons, folks usually use AIC or BIC. 

>
>2- How can I obtain the fitted values for my model?

Look at program options?

>
>3- How can I plot my mixed effects model in a way that it also shows confidence interval. I plotted the predicted value versus my independent variable but it looks like the linear line and it does not show me the confidence interval.
>
>Thanks so much for your help


There is more activity in the SPSS mailing list at SPSSX-L.
I think there are versions of the same question, which you
might find by searching the list at Nabble -- even without 
joining. 

For info about joining SPSSX-L, go to 
http://spssx-discussion.1045642.n5.nabble.com/ 

and click on the More Options link near the top of the page.


0
Rich
8/23/2016 10:53:10 PM
On Tuesday, August 23, 2016 at 6:53:14 PM UTC-4, Rich Ulrich wrote:
> On Tue, 23 Aug 2016 00:43:24 -0700 (PDT), Mojy <mgmoghaddam@gmail.com>
> wrote:
> 
> >Hello,
> >
> >I am conducting the linear mixed effects analysis by SPSS. I have few questions and highly appreciate your help:
> >
> >1- How can I compare two linear mixed effects models to find the best model?
> >I've been told to use the t-test for coefficients but how? What is the SPSS command?
> >
> >I need to compare one simple mixed effects model with the Univariate one.
> 
> For a nested comparison, you have an F test from the 
> change in mean-square and change in d.f.  If you are using
> regression, that is among the options -- after entering a Step. 

I suspect the OP is using the MIXED procedure, in which case it will be a likelihood ratio test (on the change in -2LL) rather than an F-test on the change in R^2.

> 
> For non-nested comparisons, folks usually use AIC or BIC. 

Agreed.

> 
> >
> >2- How can I obtain the fitted values for my model?
> 
> Look at program options?
> 
> >
> >3- How can I plot my mixed effects model in a way that it also shows confidence interval. I plotted the predicted value versus my independent variable but it looks like the linear line and it does not show me the confidence interval.
> >
> >Thanks so much for your help
> 
> 
> There is more activity in the SPSS mailing list at SPSSX-L.
> I think there are versions of the same question, which you
> might find by searching the list at Nabble -- even without 
> joining. 
> 
> For info about joining SPSSX-L, go to 
> http://spssx-discussion.1045642.n5.nabble.com/ 
> 
> and click on the More Options link near the top of the page.

Yes, that mailing list is far more active than this newsgroup these days.  

HTH.
0
Bruce
8/24/2016 5:09:28 PM
Reply: