1
$\begingroup$

Disclaimer: I am a coder -- not at all a research scientist, data scientist, or statistician.

I support an online survey application that is then used to generate aggregate and individual data reports and visualizations. For the sake of internal bookkeeping, we categorize all of our variables according to type (categorical, ratio, numeric, etc). We don't really gather qualitative data (and I have to admit, I'm not familiar with any ways these might be categorized).

We also collect a lot of data, that for want of a better term, we call nonresearch data -- e.g. person names, organizational names, and address information. I'm wondering if there's a better term that we could be using when talking about (and more importantly, categorizing) this data.

$\endgroup$
4
  • $\begingroup$ You're free to call your data whatever you like, especially for "internal bookkeeping." If you're looking for something "better," then it would seem you have some additional purpose in mind. Could you please explain what this purpose might be, so that we can understand what you mean by "better"? $\endgroup$
    – whuber
    Apr 28 '15 at 14:45
  • 2
    $\begingroup$ Good question. In fact, we have some data variables which we're able to sanitize so they can be used by the public for performing research queries -- these are classified as "research variables" and include categorical and numeric variables. We also have a lot of categorical and numeric variables that are obscure enough or dirty enough that we don't want them available for research. However, they're not classified as nonresearch because they're still quantifiable in some way (whereas address information is not). I'm trying to avoid confusion where we have term collisions. $\endgroup$
    – tobylaroni
    Apr 28 '15 at 14:55
  • $\begingroup$ What you call nonresearch data are often termed metadata --- that is, data about the data. $\endgroup$ Apr 28 '15 at 17:14
  • 1
    $\begingroup$ @kjetilbhalvorsen I'm a little hesitant to call this metadata since this represents information that's captured along-side the quantitative data. I feel like if these were labels used to classify a data-set, I would feel more comfortable calling it metadata. $\endgroup$
    – tobylaroni
    Apr 28 '15 at 18:07
2
$\begingroup$

In principle, this is categorical (also called "nominal") data. In a four-data-type system, that's the catch-all type for unordered, non-numerical data.

But you've unfortunately found the limit of the four-type classification system. The term "categorical" implies that the data can fall into a relatively small number of categories and follows a categorical distribution. That's technically also true for addresses, but isn't a helpful mental model. And if you're developing automated data-cleaning procedures that operate based on assigned data type, you might run into problems dealing with a large number of unique and unusually long "category" values like "4 Privet Drive, Little Whinging, Surrey, England."

I don't think there's one accepted term for this, but if you're just looking for a name for this type of category I'd consider terms like "administrative," or your own term "nonresearch." You could also just call it "qualitative" in order to distinguish it from the more-but-not-entirely-quantitative "categorical/nominal" type. You could also call it "metadata," but I feel like that would be confusing: is it "meta" with respect to observations, or "meta" with respect to the data set?

$\endgroup$
5
  • 1
    $\begingroup$ Helpful; and I like Harry Potter too. But (1) I don't agree that "categorical" implies a small number of categories, even as a first approximation. It's just that variables with few categories are the easiest and most likely to feature in statistical analysis. In practice any variables that act as identifiers (names, addresses, etc.) are likely to have many, many categories. (2) Categorical strict sense = nominal; categorical wide sense = nominal + ordinal; both uses are very common in literature I see. "Identifier variables" seems the simplest term here. $\endgroup$
    – Nick Cox
    Apr 28 '15 at 14:47
  • $\begingroup$ We used to call all non-numeric data categorical but then switched to calling it nonresearch because we could actually perform analysis on some variables (religious denomination), but you can't really do the same with "123 Cherry Lane" (or at least, not for what we're trying to do). Technically, we would call a zip code nonresearch also, but in truth it probably really is categorical (we just don't use it this way). Thanks for introducing the four-type classification system -- it does appear that this data would be nominal, although I also like the administrative term. $\endgroup$
    – tobylaroni
    Apr 28 '15 at 14:48
  • 3
    $\begingroup$ tobylaroni, since you seem unfamiliar with this "four type ... system," you would greatly benefit from--and likely enjoy--reading two short, accessible papers on the subject. The first is Stevens' original paper, On the Theory of Scales of Measurement (1946). The second is Velleman and Wilkinson's (1993) reaction, Nominal, Ordinal, Interval, and Ratio Typologies are Misleading. Both are freely accessible on the Web; search for Stevens typology. Or see stats.stackexchange.com/questions/23200 on this site. $\endgroup$
    – whuber
    Apr 28 '15 at 15:03
  • $\begingroup$ @whuber it seemed like that's what he was using, but it sounds like he came up with it independently $\endgroup$ Apr 28 '15 at 15:10
  • $\begingroup$ @NickCox that's what I think I was trying to say. $\endgroup$ Apr 28 '15 at 15:10

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.