dtburk
commented
3 years ago That's an interesting idea, thanks!
One immediate question I have is how to best handle variables for which not all the values are labelled, and the user asks read_ipums_micro() to output_labelled_as "label" or "factor".
For example, say you have an income variable where the only labelled value is 9999999, which indicates a "Missing" value. If the user requests output_labelled_as "label" or "factor", I wouldn't really want to coerce all those unlabelled numeric values to characters or factor levels.
Perhaps the best way to handle that situation would be to skip variables for which only a subset of values are labelled, with a warning, and leave them as haven-labelled vectors. Does that seem sensible?
schmert
Haven-labelled variables are unfamiliar to many R users. The {ipumsr} documentation even includes instructions that suggest that R users will almost always want to alter the haven-labelled variables output by read_ipums_micro before doing any real work -- with zap_values, to_character, etc.
Would it be possible to add a parameter to read_ipums_micro that allows the user to choose how labelled variables are output in the first place? For example, output_labelled_as = c("haven", "value", "label", "factor") with the default being the current "haven"?
This could save R users a ton of headaches. Thanks.