Assuming that you are not studying the matter of gender self-identification itself (i.e. the dynamics of the count of the very small number of people who do not consider themselves either male or female), it seems to me that a survey over time should tally (self-identified) MALE / FEMALE / DECLINE TO RESPOND. The "Decline to Respond" category will be extremely significant -- a far bigger threat to generation of an accurate count than the loss of the fraction of 1% of Wikipedians who consider themselves neither here nor there on the matter of gender. This is, of course, not particularly "sensitive to the many categories of non-male and non-female genders that exist in Wikipedia," which you apparently presume to exist a priori. Tim Davenport "Carrite" on WP /// "Randy from Boise" on WPO Corvallis, OR