Skip to content

Few details about the mmrm package #18

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
danielinteractive wants to merge 1 commit into from
Closed

Few details about the mmrm package #18

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
20 changes: 12 additions & 8 deletions src/02h_mmrm.qmd
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -600,12 +600,12 @@ lmer(data=simulated_data,
```
Note that the choice of the particular optimizer via
`control = lmerControl(optimizer="nlminbwrap")` was chosen based on trying
different ones to resolve convergence warnings. This process, as well as other
tricks (like scaling and centering covariates) are conveniently wrapped in the
dedicated
[`mmrm` `R` package](https://cran.r-project.org/web/packages/mmrm/index.html).
For this reason the `mmrm` package is preferable and its usage is demonstrated
below:
different ones to resolve convergence warnings.

Unfortunately, Daniel found out later that the `lme4` package route does not allow the fitting of MMRM with more visits, as it can easily lead to very long computation times and low convergence rates. Therefore, he started prototyping the dedicated [`mmrm` `R` package](https://cran.r-project.org/package=mmrm), which was then quickly adopted as the first open source project of the [`openstatsware`](https://openstatsware.org) group. The `mmrm` package is a completely new implementation of the MMRM, and implements the log-likelihood function in C++ with the help of the [`TMB` package]((https://cran.r-project.org/package=TMB)), which allows for automatic differentiation. It is similar with regards to the architecture to the [`glmmTMB` package](https://cran.r-project.org/package=glmmTMB), but it provides the important Satterthwaite and Kenward-Roger degrees of freedom methods. More details are available e.g. in the [R/Pharma workshop by Daniel and Doug Kelkhoff](https://rinpharma.com/publication/rinpharma_327/).

For this reason the `mmrm` package is preferable and its usage is demonstrated below:

```{r mmrm_package_example}
mmrm_fit <- mmrm(
formula = CHG ~ TRT01P + AVISIT + BASE + AVISIT:TRT01P + AVISIT:BASE + us(AVISIT | USUBJID),
Expand Down Expand Up @@ -735,8 +735,12 @@ non-identifability of the conditional approach.
The model formula above specifies that the correlation structure is the
same across all treatment groups. This is not an assumption we have to
make and SAS provides the `GROUP=TRT01P` option in the `REPEATED`
statement to avoid it. At the time of writing this chapter, `brms`,
`lme4` and the `mmrm` package do not have an option for allowing
statement to avoid it. Similarly, the `mmrm` package allows the user to specify
group specific covariance parameter estimates
using the `us(AVISIT | TRT01P / USUBJID)` syntax
(so "group / subject" on the right hand side, see
[here](https://openpharma.github.io/mmrm/latest-tag/articles/introduction.html#grouped-covariance-structure)
for details). On the other hand, the `brms` and `lme4` packages do not have an option for allowing
separate unstructed covariance matrices for the different treatment
groups. However, `brms` offers more flexibility than the other two
packages regarding the variation at different visits and treatment groups through
Expand Down