Open nhprins opened 5 years ago
note that in pediatric PK, WT and AGE are heavily correlated. Therefore simulations across balanced distributions of both covariates will not be representative for the data the predictions were based on.
I do recognize that if the function accommodates this, this will limit the creation of 'smooth curves' across either one of the correlated covariates - I'd be fine with this.
nm.process.coveffects() does nicely integrate predictions conditional on multiple categorical covariates but it ignores mutlivariate distribution of continuous covariates.
perhaps this is due to the function expand.grid that creates a balanced dataset of the continuous covariates rather than sampling from the original distribution. I acknowledge I do not know the code well enough to be certain about this but the simulation I have been looking at for pediatric PK model with allometry (body weight) and a maturation function (age) does suggest so.