Statistical Modeling, Causal Inference, and Social Science: Discarded Ipods: a sampling problem

Discarded Ipods: a sampling problem

Carrie asks:

If by any chance you're still teaching kids to do surveys, we have a project we could REALLY use help on. . . . we'd love to have a survey of ipod users asking them how many ipods they have owned, how often they used each of them, and how long they lasted before dying. We'd then like to crunch that data to find the likelihood of the ipod dying at given intervals.

Matt writes,

Barry

Andrew, that's a problem that companies have been trying to solve for a while. While I was at (insert name of Big 3 auto company here), the best answer was to get lists of buyers, and phone/mail them. It's a lot of work, and won't catch highly mobile people (like, say, college students/recent grads).

Robert

Matt wrote:

A problem with this approach is that it would over-sample people who listen to their ipods frequently. [...] This over-sampling of ipods that are used frequently would probably lead to an overestimate of the failure rate.

Why would it do that, if "time to failure" is measured as hours of play?

James Sams

I agree that you are not oversampling any population. The questions in the survey "how many hours a week do you listen to your ipod" and "for how long have you owned it," or some variation should give you an idea of hours of play, which is the important metric anyway. If you want to convert this to "months of use by average user," well, that data is also available. Of course, the estimates of hrs/wk would be somewhat inaccurate, but i'm guessing not much more so than in some of the other questions you are likely to ask. (i've seen people w/ broken audio devices that simply didn't have their headphones plugged in completely, and other trivial problems. otoh, i've seen devices in very sad shape that are considered 'operating' but at far substandard levels)

Michael Weiksner

A long shot: approach apple and see if they will share a sample of the customer database of iPod owners with you to contact for a survey. With cooperation, you might be able to take samples from different vintages, etc. Maybe they already know the answer...

Barry

Given the problems, it might be nice to use the dreaded 'convenience sample'. Have each student ask 10 friends/acquaintances about iPod usage/breakage. Then proceed from there with analyses. The nice thing is that you've got clustering in the data, which can be demonstrated, analyzed, and accounted for.

John S.

Matt wrote: "you could sample people leaving the apple store when they purchase ipods and then follow-up with them at regular intervals."

I may be wrong, but I think most iPods are purchased at Wal-Mart or Best Buy. People who shop at Apple stores are more likely to be "Apple fans" who buy new iPods not because their old one failed, but because they want the latest model. Shoppers at Apple stores are probably more affluent as well.