13

We're showing a pie graph that would show something like 53% male, 41% female and 7% unknown.

The "unknown" category includes many reasons we can't tell: we don't know, we can't know for legal reasons, we can't know for self-imposed privacy policy reasons, our data sources don't know, etc. The reasons are quite innumerable.

This seems like a common problem so I wanted to ask if there is a common solution?

|improve this question
  • 8
    What's your reason for not just saying "unknown"? Is it because in some cases you do know but can't report it? – user56reinstatemonica8 Feb 20 '15 at 23:14
  • 8
    Other implies that they aren't male or female, which isn't true. You'd need to use unknown. – Jon Feb 21 '15 at 4:29
  • 8
    Not everybody fits in M/F categories, and for those that do fit in those categories you don't have 100% data. That's the reality of the world we live in. What is the problem? ;-) – adrianh Feb 21 '15 at 8:38
  • 5
    You can always make a clear cut by asking "Do you have a Y chromosome?". ;-) – Kos Feb 21 '15 at 12:05
  • 5
    Gender isn't necessarily binary. – DA01 Feb 22 '15 at 21:12
19

Assuming you have a context where this level of accuracy matters (e.g. an academic or technical audience, or a delicate topic) it's always best to just clearly and simply state what's going on.

This 7% have a gender (male, female, or something else). For some, you know it, for some you don't. In all cases, you're not reporting it. So just say that:

  • Not reported
  • Not published
  • Not shared
  • Not publicly available
  • Unavailable/confidential
  • Unknown/withheld
  • Unspecified/other (suggested in a comment by PLL below)

...or if your organisation has a more informal tone of voice:

  • Can't say
  • Private

...or if your organisation has an [annoyingly] chatty tone of voice:

  • ???
  • Secret
  • Wouldn't you like to know?
|improve this answer
  • 3
    "Data unavailable," "Not disclosed," "Withheld," and "Confidential" all imply facts about the 7% that are not (known to be) true for all of the 7%, and are thus bad suggestions. – KRyan Feb 21 '15 at 20:36
  • True. Have edited those ones – user56reinstatemonica8 Feb 22 '15 at 21:07
  • It's not necessarily true that those 7% have a gender. Agendered individuals exist. Just a minor quibble though, as the main thrust of your answer is a good one. – dhmstark Feb 24 '15 at 23:42
10

It seems that "Unknown" is true to the situation. Go with that. You could always explain, as you did on this site, why gender is unknown.

|improve this answer
  • 16
    Wondering if "unspecified" might be a better term in implying the responders from the studies choose not to provide an answer, as oppose to they not knowing. – nightning Feb 20 '15 at 21:57
  • Unspecified is very good, but might technically not be true in cases where it was specified but can't be reported like "legal reasons" or "self-imposed privacy policy". Whether this matters depends on the pedantry of your audience... – user56reinstatemonica8 Feb 20 '15 at 22:56
  • It's not 'truthiest' it's actualy 'true' in that they simply don't know, hence 'unknown' – DA01 Feb 22 '15 at 21:12
5

The usual third option is Other, which would encompass a variety of options including but not limited to: those who did not fall in to the Male/Female box; those that explicitly would rather not say; those who implicitly did not say; those where the data is not known, etc, etc.

And then if you feel the need to explain what 'other' actually entails, make it a link to a decent explanation that covers some of the popular reasons whilst respecting that there can be any number of other reasons not listed. That would at least show that you recognise the options rather than leaving the user to guess what Other means.

There is a third way - which is just to leave the 7% blank and not label it at all, but still that leaves the user guessing - although users aren't so dumb...

Clearly, reasons for 'would rather not say' are not the same reasons behind explicitly choosing 'other', which is one of the things that bothers me about such gender options in forms - assuming the data is really even necessary in the first place! In reality the 'would rather not say' could partially contribute to any one of Male, Female or Other, so if this is significant in your scenario, and you have the data, then you may want to rethink having just three categories.

|improve this answer
  • 14
    The issue I see with other is that implies that the person is not female or male, but other . Which is probably not the case for all the 7%. – rr1g0 Feb 20 '15 at 21:41
  • 20
    "Other" and "Unknown" are different and distinct things, mixing them up is usually misleading. Here, if I saw a pie chart with 53% male, 41% female and 7% other, my reaction would be "Crikey, this sample has a surprisingly large trans/intersex population! I wonder why that is?". Or consider "Cause of death: other" (a rare event has occurred, we know what it is) vs "Cause of death: unknown" (doctors are baffled) – user56reinstatemonica8 Feb 20 '15 at 23:21
  • 6
    If you want to cover both cases, without subdivision, you can simply mark the third slice as “other/unspecified”. – PLL Feb 20 '15 at 23:39
  • 1
    @PLL Please make that an answer; none of the existing answers have suggested that, and I think it is (by far) the best solution; I think the existing answers each have a particular problem with them. I would upvote that answer; I am not upvoting any of the others at this point. – KRyan Feb 21 '15 at 20:35
  • 1
    It depends somewhat on the label/caption of the pie chart whether other is likely to be interpreted as other answer rather than other gender. – Crissov Feb 24 '15 at 7:03
4

n/a seems to fit well, and it is a known convention.

n/a or N/A is a common abbreviation in tables and lists for not applicable, not available or no answer.

|improve this answer
  • 4
    This has a similar problem to "other" in that it sounds like it's known that gender is not applicable to 7% of respondents; when actually we expect they do have a gender, we just don't know what it is. e.g. N/A would fit respondents that couldn't have a gender (e.g. responses on behalf of an organisation), or people who don't identify with any gender (such people do exist - but it's much rarer than 7%). I'd suggest "No data" which is similar but makes it clear that it's an unknown. – user56reinstatemonica8 Feb 20 '15 at 22:52
  • 2
    n/a means more than one thing. It also means "no answer" and "not available" (as I quoted in my answer). The problem you say could happen if people don't know that about n/a... which I guess it could happen, but it still is a conventional way of showing the absence of data. Perhaps depends on who is seeing this. – rr1g0 Feb 20 '15 at 22:57
  • @rr1g0 Except that not available is the most common understanding of N/A, and is thus likely to be misunderstood. – KRyan Feb 21 '15 at 20:38
  • 1
    @KRyan I'm not sure if I see your point, but Not available is fine in this case, the problem would be with Not applicable, as user568458 noted. Not available is the same as No data. – rr1g0 Feb 21 '15 at 20:44
  • 1
    @rr1g0 Yes... I meant not applicable but wrote not available for reasons even I cannot begin to imagine. Not applicable is my usual understanding of N/A. – KRyan Feb 21 '15 at 20:46
0

You are asking for pie graphs specifically which are good for one thing and one thing only: showing which part or which combination/coalition has a share of 50% or more. Most often, they should have just two segments or visually highlighted segment groups.

You currently want to show three data points at once, but your data set is ordered hierarchically:

  • Participants whose gender we do not show.
    • Participants whose gender we do not know.
    • Participants whose gender we know but are not allowed to share.
    • Participants whose gender is not covered by our data structure – some transsexuals, most intersexuals, corporations, groups, …
  • Participants with disclosed classic gender.
    • Participants who self-assign to the female gender – women.
    • Participants who self-assign to the male gender – men.

You want to collapse the sub-points of the first bullet into one, but not of the latter. That is a categorical error!

You have three, maybe four choices of diagrams:

  • Show the top tiers only.
  • Show the reasons for not showing the gender.
  • Show fe-/male proportions only.
  • Combine the last two, i.e. show all lower tiers.

Those show very different things, not all of which are of actual interest. You obviously want the third one, or a combination of the first and the third one. That is possible, but not as linear segments within a single pie chart: You could put the 7% N/A either into an outer ring or into a center circle, probably painted grayish (or semi-transparent); the former works well with multiple pie charts shown next to each other because you get to see data and quality of data at once, the latter may be better stand-alone. Worst example of pie charts ever

To finally answer your actual question, the correct label for each segment follows automatically from correctly labeled charts.

|improve this answer
  • Down-voters, care to explain? – Crissov Jan 28 '16 at 15:49

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.