AR1 in multilevel models PreviousNext
Mplus Discussion > Multilevel Data/Complex Sample >
Message/Author
 Oliver posted on Thursday, March 17, 2016 - 12:44 pm
Linda and Bengt,

I am currently using the Mplus demo and it works perfectly. Your software seems to be fantastic and seems to offer more possibilities than SPSS, the software that I'm currently using.

In my case, I am primarily interested in conducting multilevel analyses, in which daily diary data (Level 1) are nested within participants (Level 2). For instance, in one of my diary/longitudinal studies, participants (n = 100) provided pain (Level 1) and mood (Level 1) ratings, once a day for 14 consecutive days. Similarly to the previous post from Jonathon Rendina back in 2013, I'm interested in examining the Level 1 association between pain and mood, while accounting for the autoregressive effect that repeated measurements create on the day-level measurement of the outcome. In SPSS and SAS, this is usually done by incorporating an AR1 component to the model. Peer-reviewers tend to be very critical if the AR1 is not specified, and I'm wondering whether there's a way in Mplus to incorporate the AR1. IF so, how would it be incorporate to the syntax below ? Thanks in advance for your support...

DATA: FILE = Datab2.dat;
VARIABLE: NAMES = ID L1Pain L1Mood;
WITHIN = L1Pain;
CLUSTER = ID;
DEFINE: CENTER L1Pain (GRANDMEAN);
ANALYSIS: TYPE = TWOLEVEL;
MODEL:
%WITHIN%
L1Mood ON L1Pain;
 Tihomir Asparouhov posted on Friday, March 18, 2016 - 2:23 pm
The simplest way at this point is to follow example User's Guide 6.17. You will have y1-y14 and model constraints will have 13 parameters. Also remove "S". The model will be run as single level (wide), not two level (long).

I would also recommend reading

"Conditions for the Equivalence of the Autoregressive Latent Trajectory Model and a Latent Growth Curve Model With Autoregressive Disturbances" ELLEN L. HAMAKER

Incidentally in the upcoming V8 this will be very easy and very general
http://mplus.fss.uu.nl/wp-content/uploads/sites/24/2012/07/tihomir-DSEMslides.pdf
 Oliver posted on Tuesday, March 22, 2016 - 12:19 pm
Thank you for your response, Tihomir.

I'm glad to hear that it will be possible to model the AR1 parameter more easily in the next Mplus version (V8). By any chance, do you know if V8 will allow us to model the AR1 parameter using the "long/stacked" dataset format ? It would be much easier than using the the "wide" format given that all my current databases involving longitudinal/repeated measures are currently in the long format...

Thanks in advance for letting me know,
MO.
 Bengt O. Muthen posted on Tuesday, March 22, 2016 - 1:21 pm
Yes. So as 2-level data.
 Bep Uink posted on Thursday, March 24, 2016 - 12:21 am
Hello. I was wondering if MPLUS was able to run multilevel threshold autoregressive models, as discussed in Haan-Rietdijk et al (2016) http://link.springer.com/article/10.1007/s11336-014-9417-x/fulltext.html. The authors suggest Bayesian based software analysis is currently the best for such models, but wondering if you had idea on how to model two auto regression co-efficients (upper and lower) simultaneously in MPlus. Thank you.
 Tihomir Asparouhov posted on Friday, March 25, 2016 - 8:41 am
No, but in the upcoming version 8 there will be some two-level AR models. See

http://mplus.fss.uu.nl/wp-content/uploads/sites/24/2012/07/2016JanSkypetalk.pdf

and

http://mplus.fss.uu.nl/wp-content/uploads/sites/24/2012/07/tihomir-DSEMslides.pdf
Back to top
Add Your Message Here
Post:
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Password:
Options: Enable HTML code in message
Automatically activate URLs in message
Action: