Closed cjvanlissa closed 3 years ago
I imagine it could be helpful to call closed_data()
without synthesizing if-then load_data()
would issue a message about how to obtain the real data. Or maybe we build a new function that does that (called private_data()
or requestable_data()
)?
I agree. I ran into the following issues recently: 1) A lavaan model did not converge on the synthetic data, thus rendering the remainder of my code non-reproducible. The problem was solved by simulating data from the lavaan model directly - but there was no function to replace the synthetic_data.csv 2) Data were too large to be synthesized using random forests. Simple bootstrapping per column would work, and synthetic() allows for this - but closed_data() does not allow users to modify the call to synthetic()
Ah I didn't read carefully enough! Couldn't you pass the dots from closed_data
to save_data
and from there to synthetic()
?
I would do this for you, but I do not understand the magic of all_args()
@aaronpeikert I added all_args() to avoid problems with list(match.call()[-1]) where, for example, in tests, functions would fail because the passed object were within some higher parent environment and not within scope for the final function down the call stack.
Should it be possible to call closed_data() without synthesizing data? It should definitely be possible to manually add a synthetic dataset.