Integration several datasets about ibis.isdm HOT 5 CLOSED

Not quite sure what you mean, but passing on predictions to the next models is the default for engines that don't support joint estimation of multiple datasets, the parameter method_integration applies (default: "predictor"). Extracted predictors are in this case added.

• It needs to pass on predictions among multiple steps. One can think of an example where like 3 or more different datasets are added.
• Ultimately the number of predictors in the model object does not matter as long as they are present. Spatial (or temporal) projections only make use of those predictors in the equation (formula).

from ibis.isdm.

• It needs to pass on predictions among multiple steps. One can think of an example where like 3 or more different datasets are added.
• Ultimately the number of predictors in the model object does not matter as long as they are present. Spatial (or temporal) projections only make use of those predictors in the equation (formula).

I understand that and it's done within the loop in train() as well as updating the formula object. My question is why in the code example above the extracted predictor values are added to the first data.frame which stores covariates a + b. Doesn't it only require c + d+ prev.prediction but NOT a + b + prev.prediction? For example, the first formula is also not updated using a skip, but the data.frame still is.

Even if its not particular used during any later predictions, it adds unnecessary to the data.

from ibis.isdm.

My question is why in the code example above the extracted predictor values are added to the first data.frame which stores covariates a + b. Doesn't it only require c + d+ prev.prediction but NOT a + b + prev.prediction?

If at some point we want project() to reproduce the formula of train() we would need all the predictors in the dataset (unless full models are saved in the object, which would be even larger in size than additional columns).

Even if its not particular used during any later predictions, it adds unnecessary to the data.

if you think it is critical to fix, and it does not affect the functionality of method_integration="predictor" , then feel free to commit a fix.

from ibis.isdm.

I think this can be closed due to #119 right?

from ibis.isdm.

I think so yes

from ibis.isdm.