Currently we have that imputation of d/p/q/r should be wrappers, for example for a distribution X with a pdf and no cdf, a user can impute the cdf via:
ImputeCdf$new(X, strategy = "pdf2cdf")
However the problem with this is that if a user imputes p/q/r all from d then they end up with a distribution called imputedR_imputedQ_imputedP_X with parameters imputedR_imputedQ_imputedP_param1.
I think it is more sensible to have these imputations as functions that take as input the distribution and alter the respective function. e.g. imputeCDF(X) would return X but with the cdf imputed via a given strategy. Then we could include dynamic comments in the function and add to the distribution description a note detailing the imputation.
RaphaelS1
commented
5 years ago
Currently we have that imputation of d/p/q/r should be wrappers, for example for a distribution X with a pdf and no cdf, a user can impute the cdf via:
However the problem with this is that if a user imputes p/q/r all from d then they end up with a distribution called
imputedR_imputedQ_imputedP_Xwith parametersimputedR_imputedQ_imputedP_param1.I think it is more sensible to have these imputations as functions that take as input the distribution and alter the respective function. e.g. imputeCDF(X) would return X but with the cdf imputed via a given strategy. Then we could include dynamic comments in the function and add to the distribution description a note detailing the imputation.
Thoughts?