At this moment we are storing our reproducible result of a statistical analysis (e.g. LMM) as an R object which contains input data, parameters, models and output.
We use the serializeJSON function from the JSONlite package to create a very verbose JSON file which can be converted back to an .Rdata object by the unserializeJSON function
Store the complex object as .Rdata file and create plumber api’s to extract the data for non R applications.
In situation 1, the stored file can directly be used by other, non R, applications. But is this situation sustainable, which solution would provide the best compatibility for - future R versions - more complex and - larger objects?
Is there anyone who could share some experience or ideas about the best practice?