Slashdot Mirror

← Back to Stories (view on slashdot.org)

A Broadband Survey That Asks the Right Questions

Posted by timothy on from the how-much-hentai-is-enough? dept.

Lauren Weinstein writes "I've just deployed the first ever Broadband Survey under the auspices of GCTIP, which asks questions that the FCC neglected to ask about service types, promised vs. actual broadband speeds, user satisfaction (or lack thereof) with their ISPs and local ISP competition, etc. I'm already finding the detailed comments many persons are leaving on the survey form to be extremely illuminating and with sufficient participation I'm hoping my reports from this data will be useful to the Internet community broadly."

1 of 120 comments (clear)

  1. I teach survey design... This is terrible. by kklein · · Score: 5, Insightful

    This, in all honesty, is the worst survey I have ever seen, and I work with language teachers.

    Not only are you setting yourself up for selection bias (as many others have pointed out), you've got all these free-answer text boxes all over it. Have you given any thought whatsoever to what you're going to do with the "data" that you get from this instrument?

    Things like network speed should be in set categories. Satisfaction should be on a Likert scale, and should be broken down into aspects of interest (satisfaction with upload, download, etc.). The ISPs should be on a drop-down menu, not free answer (you'll need to include an "Other"). ZIP and City should be in separate fields (how are you going to parse those?--yes, it can be done, BUT WHY???).

    Your question about maximum upload and download speed and limit and favorite color... Son, you make me want to stab out my eyes with a fork. What are you asking with that question? Whatever it is, it should be several questions with constrained responses.

    One of the cardinal rules of survey design is that it should be quick and easy for people to fill out. Do the hard work for them, and let them just tick boxes. If you don't, they won't take it and all you'll get is data skewed toward people who--like you--actually care enough to type up a bunch of thoughts. I care about broadband, but even I am not interested in blathering away into a text box.

    Pray tell, what "statistical purposes" would my email address be used for? Last I knew, principal components analysis only took numeric data... Same for cluster analysis. "This will only be used for magical statistics that use email addresses as variables... Or if we want to drop a line and say hi." Please.

    You are setting yourself up for a world of hurt. You will need to go through with Nvivo or something to categorize all the garbage you get from this, and even if you present results, all you're really going to be presenting is "here is some stuff that people said." I have no time for listening to results of surveys like that. It's softheaded gibberish.

    You are lucky you're not a student in my research practicum. There's no way I'd sign off on this as a research instrument.