Remind the novice user of labelled-format?

[sda030]

I do not know how many times R-novice colleagues have misunderstood the labelled-class vs. factor-distinction when importing Stata/SPSS-data. In our organization and field, custom missing types are very rare and obscure, so I can safely tell them to always use labelled::unlabelled(), which they of course find counter-intuitive (they don't want to remove labels! :) ). This issue might even trip up the experienced R-user who for the first time has to import a Stata/SPSS dataset and jumps into regression-analyses.

I understand the need for labelled-class, but could we perhaps generate a one-liner warning/notification? Perhaps something like this IF there are no multiple NAs across the categorical variables.

cli::cli_inform(c(i="You likely want to convert categorical variabels from {.cls labelled}-class to {.cls factor} before data-analyses."))
#> ℹ You likely want to convert categorical variabels from <labelled>-class to <factor> before data-analyses.

It could be suppressed with an argument that defaults to FALSE. And perhaps also respecting a option("haven.quiet_labelled_notification" = TRUE) for the more advanced users.

[gorcha]

Hi @sda030,

Thanks for the feedback!

I'm reluctant to add a message like this to the package since it would be quite noisy and it's a fundamental part of the package that is mentioned prominently in the README.

Having said that, it could definitely be more clearly noted in the documentation for the read_*() functions. I think we're due for a general documentation cleanup, so I'll keep this in mind as part of that.

[sda030]

Hi @gorcha, that is understandable. We found another solution with the sjlabelled::read_*.