Hello,
I am working with some longitudinal data resulted from a clinical trial. I have the interaction effect of time and group + random effect of Subject. Time variable consists of 6 timepoint and group has treatment and control.
First, in order to be able to see the change from one timepoint to another, I created a factor version of the time and group variables, and fitted the model as following:
lmer("Depression~Time_factor*group_factor+(1| Subject)"), data=data)
I would be ok with this, but the output gives a single intercept for both groups, which is just the mean for treatment group at timepoint 1, and a beta for each timepoint per group
which is basically the difference between the intercept and a certain group. For example, the mean for timepoint 1 control group is expreced as a beta, or the difference between the mean of treatment group timepoint 1 and control group timepoint 1.
Another option was to fit a separate model for each group with the function by.
by(data=data, INDICES=data$Group, FUN=lmer,
formula=formula("Depression~Time_factor+(1|Subject)"))
Both seem to give appropriate information regarding within-group change, but I am not sure which one to use, and which one is the correct one.
Thank you for your answers. I hope I explained everything clear enough.
Have a nice day.
See the course notes of one of the developers of these ideas used to follow the effect of drug cocktails on HI V infection, Prof. M. Davidian:
https://www4.stat.ncsu.edu/~davidian/st732/notes.html
Best, D. Booth
Dear Stefan Marian ,
from what I understand, you are interested in the interaction of your between-groups variable with the within-groups variable, only. Additionally you want to explore the within-group effects on your dependent variable by between-group levels (?). Or are you solely interested in the effects of your within groups variable? You should answer these questions before modelling because they affect how your model should look like.
In any case I think you are looking for something like the emmeans-package. Among other tools, it also includes the emmeans-function which allows you to further explore your model in terms of comparison of group means by factor levels. This should also help understanding the model summary. Try:
model_1
It seems to me that your data is made up of records that are made up of the trajectories of depression over times 1-6 for two treatment groups (treated and not-treated)? I would not view Time as a factor acting on Depression. I would treat the string of Depression observations over Time as a vector of correlated observations. An observation record would look to me like:
(d1, d2, d3, d4, d5, d6)
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Software
- R