Open clarkliming opened 9 months ago
related to #387 but still a little different
in some scenarios the variance estimate is 0 if the values match so well to the fixed effect, leading to a incorrect hessian/gradient
see the following example
fit <- mmrm(FEV1 ~ RACE + SEX + ar1(AVISIT | USUBJID), data = fev_data) fev_data$FEV1[!is.na(fev_data$FEV1)] <- fitted(fit) mmrm(FEV1 ~ RACE + SEX + ar1(AVISIT | USUBJID), data = fev_data)
we might improve this error messages to be clearer
related to #387 but still a little different
in some scenarios the variance estimate is 0 if the values match so well to the fixed effect, leading to a incorrect hessian/gradient
see the following example
we might improve this error messages to be clearer