is this formula wrong here?

[clarkliming]

mmrm/simulations/missing-data-benchmarks/R/method/nlme-wrapper.R

Line 59 in b0741f8

correlation = nlme::corSymm(form = ~ 1 | participant),

I belive we should use the ~ visit|id as the covariance structure in nlme, what do you think? @danielinteractive

[danielinteractive]

Thanks @clarkliming , good question - from the doc of corSymm it says "Defaults to ~ 1, which corresponds to using the order of the observations in the data as a covariate, and no groups.", so if the order is correct, then it should be ok incidentally. But I guess to be more safe, ~ visit | id is better.

[clarkliming]

~1|id would only work if

1. for unstructured covariance structure, the first m consecutive visits are collected
2. for toeplitz, consecutive visits are collected
3. for compound symmtry, does not have any impact

[danielinteractive]

ok thanks @clarkliming , then definitely better to change it, thanks again for finding this!