Hello! I’m reading 2 journal articles that assess change/growth over time (longitudinal study), yet report using different stats by name (but I suspect they are similar if not the same). Can anyone provide me with clarity whether the statistics used in the articles are the same/different? I am not too familiar with multi-level modelling.
The first article uses a growth curve analysis to identify the score stability/growth trajectory of four composite score on a neuropsychological test across repeated test administrations, while allowing time to vary by individual. They use Mplus 7.31 to do the analysis. A random-effects model was used.
The second article assesses the pattern of change among test intervals for a neurocognitive test. “For longitudinal analysis of changes in NP testing, a repeated measures mixed model was examined among participants with two or more assessment time points using Proc Mixed (SAS 9.3). In order to account for differences in individual cognitive trajectories, individual participant improvements were modelled as a random effect, with the time effect modelled as a fixed effect.”
Thanks in advance!
They sound that they are both repeated measures random effects models;
There is quite a lot of support for you here
http://www.bristol.ac.uk/cmm/software/mlwin/mlwin-resources.html#multipleresp
see repeated measures
I think it would help you a lot if you followed a step by step analysis, such as
http://www.bristol.ac.uk/media-library/sites/cmm/migrated/documents/repeated-measures.pdf
This is also a very good site with lots of useful material
http://www.pilesofvariance.com/
They sound that they are both repeated measures random effects models;
There is quite a lot of support for you here
http://www.bristol.ac.uk/cmm/software/mlwin/mlwin-resources.html#multipleresp
see repeated measures
I think it would help you a lot if you followed a step by step analysis, such as
http://www.bristol.ac.uk/media-library/sites/cmm/migrated/documents/repeated-measures.pdf
This is also a very good site with lots of useful material
http://www.pilesofvariance.com/
Indeed, the terminology can be very confusing, particularly because somtimes different terms can mean something similar (multilevel vs. mixed). On the other hand, for example, the term "growth curve analysis" can refer to a multilevel model, but also a structural equation model (although, under certain condisitions, they are identical).
I am not sure if this helps, but chapter 16 in the "Multilevel Analysis" book by Hox, Moerbeek, & van de Schoot gives a nice comparison.
So does this video: https://www.youtube.com/watch?v=hC84mK1IYBM
The advantage of multilevel random effects over SEM is that you can model continuous time (decimal point); this can give more precise results eg if you know it is a 3.2 year old child not just 3.
The comparison is done in this paper
Multilevel models for longitudinal data Fiona Steele
http://onlinelibrary.wiley.com/doi/10.1111/j.1467-985X.2007.00509.x/abstract
Thanks very much for the advice & readings etc. They have provided me with some clarity. Much appreciated.
Growth curve analyses provide a simple and effective way of analyzing longitudinal data in such a way that the investigator can understand what they are doing without becoming proficient as to the many assumptions underlying mixed models, which do include elements of growth curve analyses.
- Badges
- Science topic
- Similar topics
- Journal Articles