Hello,
I am using the nlme R package to perform mixed models in the following experimental design:
- Outcome: Continuous variable measured in two points of time (baseline and follow-up).
- Predictor: Continuous variable measured at both visits (baseline and follow-up).
- Visit= Dummy variable: Baseline or Follow-up (0 vs 1)
- ID: Case ID (1 to n)
So, my first step was to reshape my data frame to "long" format. Hence, the "Outcome" and "Predictor" variables contain information of the baseline and follow-up visit. I subsequently determined which random effects fitted better to my outcome with the following code:
model1
I have a different take on this as I think you do not need a mixed model as you only have two time points - a pre -test and a post - the model can be done effectively by a standard multiple regression
I usually prefer Post score regressed on Pre score rather than a change score - with the latter a change of 10 from a base of 0 is very different from a change of 10 from a base of a 1000
Gelman and Hill explain this in more detail: .
Gelman, A., and J. Hill. 2006. Data Analysis Using Regression and Multilevel/Hierarchical Models. Cambridge University Press. www.stat.columbia.edu/~gelman/arm/
In preference I would explicitly model the change choosing the value for the later visit as the response and the values for the earlier visit as a predictor and then including other predictors - including any treatments and covariates (usually from the first visit and any time invariant ones). By setting up the model in this way you are conditioning on the previous starting values and explicitly modeling change.
Typically I would center the values for the first visit around their median and then B0 will be the value predicted for the second visit for the typical patient ( depending on what other predictors and how they have been centred).. I would also interact the first visit variable with other predictors to see for example if men improved more than women.
A quite complex example of this will be found in Appendix 1: Phase 1 modelling – an illustrative sequence .........- it is more complex because it is a two level model when pupils are at level 1 and the intervention is at level 2 - the class. ( the complexity does not come from the two occasions )
http://dera.ioe.ac.uk/21564/1/whiteboards_expansion.pdf
but the concepts are the same.
Article Evaluation of the primary schools whiteboard expansion project
Significant coefficients should be interpreted directly.
Variance com0nents are also interprtable.
Hope this helps,
Matthew
I have a different take on this as I think you do not need a mixed model as you only have two time points - a pre -test and a post - the model can be done effectively by a standard multiple regression
I usually prefer Post score regressed on Pre score rather than a change score - with the latter a change of 10 from a base of 0 is very different from a change of 10 from a base of a 1000
Gelman and Hill explain this in more detail: .
Gelman, A., and J. Hill. 2006. Data Analysis Using Regression and Multilevel/Hierarchical Models. Cambridge University Press. www.stat.columbia.edu/~gelman/arm/
In preference I would explicitly model the change choosing the value for the later visit as the response and the values for the earlier visit as a predictor and then including other predictors - including any treatments and covariates (usually from the first visit and any time invariant ones). By setting up the model in this way you are conditioning on the previous starting values and explicitly modeling change.
Typically I would center the values for the first visit around their median and then B0 will be the value predicted for the second visit for the typical patient ( depending on what other predictors and how they have been centred).. I would also interact the first visit variable with other predictors to see for example if men improved more than women.
A quite complex example of this will be found in Appendix 1: Phase 1 modelling – an illustrative sequence .........- it is more complex because it is a two level model when pupils are at level 1 and the intervention is at level 2 - the class. ( the complexity does not come from the two occasions )
http://dera.ioe.ac.uk/21564/1/whiteboards_expansion.pdf
but the concepts are the same.
Article Evaluation of the primary schools whiteboard expansion project
Your interpretation of coefficients is correct.
If I understand correctly, you only have one measure per visit. You should therefore not use both predictors (Predictor and visit) simultaneously. This is also not enough data on participant-level for a multilevel model.
- Badges
- Science topic
- Similar topics
- Philosophy
- Esthetics
- Interpretation