Joy of Stats Book Cover
Photo by bourgeoisbee / CC BY-NC

Joy of Stats Book Cover

I have been working recently on importing data into SPSS from the first part of my April survey on World of Warcraft motivations. This has been a fairly straightforward process for the most part. The exception is the last question about the respondent’s guild type and purpose. The question was presented as the following:

In a short sentence (140 characters), describe the primary purpose of the guild in which you spend most of your time, or enter “no guild”.

Example: I’m in a social guild that believes in random acts of kindness. We love to dance but we also raid end-game content with other casual guilds.

I am not happy with how the coding for the type and purpose is going. When I created the SPSS codebook for that part of the survey, I initially broke it down into two parts: a type and a primary purpose. The type represented a breakdown primarily between “social” and “hardcore”. The purpose could be a key activity or a purpose. The divisions were chosen based on an an initial review of the received responses (see Table 1).

Variable SPSS Variable Coding Instructions
Primary purpose of guild where most time spent

Person.TimeConsuming.
GuildPurpose.Type

0 = Not codable
1 = No guild
2 = Social guild
3 = Hardcore guild

Person.TimeConsuming.
GuildPurpose.Activity

0 = Not codable
1 = No guild
2 = Banking
3 = Role-playing
4 = Raiding
5 = Levelling
6 = Fun
99 = No identifiable activity

Table 1: Codebook Attempt #1 for Guild Type and Purpose

The “No guild” choice is self-evident. “Not codable” was intended for responses like “Forks” or “I am an herbalist in my guild”. This was usually where the respondent had not answered the question correctly, so no type or purpose could be determined. “No identifiable activity” was intended for responses like “Social”, where no obvious activity was ascertainable.

Andrew and I independently coded the 51 responses according to type and activity. We were allowed to use general knowledge about World of Warcraft but not specific knowledge about any guilds or people if identifiable somehow from the response. The response itself had to dictate the type and purpose.

We almost completely agreed on “type”, but we disagreed by about 28% on the “activity”. If an answer mentioned raiding and that seemed more key than anything else included (or nothing else was included), I coded it as “raiding”. Andrew often coded it as “fun” because he felt the raiding was deprecated or a lesser part even if it was the only thing mentioned. So, for example, if someone said something like “I’m in a social guild and we also do some raiding”, Andrew coded that as “fun”, because the “also raid” was a lesser thing, whereas I would have said “raiding”.

To fix that, he suggested that we add a category that indicated the amount of raiding. I would then have “raiding” and “some raiding”. That seemed sensible, but then left the question of what the “primary” purpose of the guild was if “some raiding” was ancillary. While the survey specifically asked for the primary purpose, but many of the responses, likely because of the poorly chosen example, included one or more activities. When coding, that meant the coder had to pick the one activity that seemed most prominent.

A colleague, in discussion about the first attempt at coding, suggested going for as wide a range of activities as possible before cutting back. I had, for example, initially omitted “performing random acts of kindness”, which occurred fairly often, as an activity. She thought that should be added as well as “social.” Social” as an activity was omitted because I was using the “social” designator appearing in responses to code for the guild type. However, in retrospect, it occurred to me it would be even more appropriate if I changed the guild type to be as interpreted by me based on my assessment of the respondent’s answer; the question did not ask the respondent to specify their guild type but to describe the purpose. I need to stay true to the question’s intent.

A modified codebook then would look something like Table 2.

Variable SPSS Variable Coding Instructions
Primary purpose of guild where most time spent Person.TimeConsuming.
GuildPurpose.Type

0 = Not codable
1 = No guild
2 = Social guild
3 = Casual guild
4 = Hardcore guild

Person.TimeConsuming.
GuildPurpose.Activity

0 = Not codable
1 = No guild
2 = Banking
3 = Levelling
4 = Philanthropy
5 = Fun
6 = Role-playing
7 = Social
8 = Some raiding
8 = Raiding
99 = No identifiable activity

Table 2: Codebook Attempt #2 for Guild Type and Purpose

That does mean that I need more formal definitions of “social”, “hardcore”, and “casual” then to use for the coding guild type. It also means there is a question about how many activities to code, given that very few responses only mentioned one purpose or activity. While the question did ask for a primary activity, I do not have a problem per se with coding for more activities. The questions are then: how many and should order be significant?

Response
  • The guild I am part of is primarily a social guild but we also get involved in raiding.
  • It a social guild with occasional raids
  • Roleplaying, adventuring, helping each other and gernally [sic] having a good time
  • [guild name] As Crafter
  • I’m in a social guild. We have a great sense of humor
  • I’m in a social guild that believes in random acts of kindness. We love to dance but we also raid end-game content with other casual guilds.
Table 3: Some sample responses for guild purpose.

Open Issues for Comment and Resolution

  • Is the second draft draft more sane in terms of how type and activity are determined in general?
  • What are some existing formal definitions of guild types that can be used?
  • How many purposes or activities should be coded?
  • If multiple activities or purposes are coded, is the order significant? That is, should I assume that the first thing mentioned is the most important, the second, the next, and so on?

Any opinions or help are appreciated, as this type of analysis is new to me.

No related posts.