Questions about the Convert Columns Dialog

[rachelkg]

Hello @rdstern, @N-thony and @Patowhiz,

I have been working on the help for the Prepare > Data Frame > Convert Columns dialog.

I am confused by the checkboxes at the bottom and how they interact.

By default Keep Attributes and Create Labels are checked. When I look it up Keep Attributes when converting variables in R online the description is as follows. "This function converts a variable into a factor, but preserves variable and value label attributes." This makes sense.

However, I seem to get the same result regardless of what I have checked. I tried Keep Attributes and Ignore Labels, but when I look at Level/Labels It looks exactly the same as when I have Create Label's Checked. Furthermore, When I uncheck Keep Attributes and check Ignore Labels I still have the same level's and labels results. I therefor feel that I must be misunderstanding thes options and how they interact with each other.

Please could someone explain.

What do these items mean?

• Keep attributes
• Ignore Labels
• Create Labels
  How to they interact with each other?
  And Finally, ignore Labels and Create Labels seem to be mutually exclusive, if so should they be radio buttons instead of check boxes?

All Insight welcome.

Thanks,

Rachel

Thank you,

Rachel

[rachelkg]

@rdstern has responded with this explanation:

That's a messy part of R-Instat, because we are trying to give you the "best of both worlds". R uses factors for categorical data, while other (still popular) statistics packages uses the idea of labelled data. Labels in SPSS are similar to the factor labels in R, but not quite the same. The attributes of a variable are its properties - that's what you see when you look at the column metadata.

There are two positive developments in R-Instat that help with this dialog. (I'm writing as you might put in the help?). The first is that you now don't need to use it very much. That's because we've improved the right-click options for converting the data a lot. So where possible we suggest you use those. The second, is that, as from Version 0.8 we have added a spreadsheet-style undo. So, now, you can simply try the conversion and if it isn't exactly what you want, then you can undo!

The complications in convert are mainly when you convert a variable into a numeric column. What do you want to do with any observations that are not numbers? Would you like to be able to convert back again, say to a factor, or character column, where they are fine? If you would like the option to convert back, then labelling (or keeping the labels) is the probably the way to go. You should link in the help to the popup dialog we have for converting to numeric on the right-click! It's there, but blank, of course. But I'm still glass-half-full, because it's nice we at least have it there!