christophergandrud / coresim Goto Github PK

dat <- tibble(x.x = 1:10,
              y = rnorm(x.x)) %>%
  as.data.frame()

mod <- lm(y ~ x.x, data = dat)

a <- qi_builder(mod,
                newdata = dat,
                slim = FALSE,
                ci = 1
                )

Causes the error

Error in possible_interaction_terms(fitted_names, n = k) : 
  Number of possible interaction terms cannot exceed number of base terms.

If I replace the . with a _, the error goes away.

In qi_builder if slim = FALSE and we get returns of simulated values, can we also get a column stating which sim a value came from? This would 1) make it easier to use the group argument in a ggplot to, say, make a bunch of densities of predicted values to compare to the distribution of the data and 2) mean that we could tell more easily when different simulated predictions came from different initial data points (e.g., if X1 = X2, if they're from the same simulation, we know they are different data points).

Add the ability to automatically detect and simulate for variables that were created with splines.

If we are getting raw simulation output, can the column name of the output simulations be the name of the response variable rather than qi_ - otherwise the user has to go through that step themselves.

Add ability to work with models that have categorical dependent variables.