Study refutes 'Net geeks' repute | CNET News.com There is something strange about finding references to data that is nearly four years old(!!) as something new and refreshing. The above-mentioned article refers to data collected by the UCLA World Internet project lead by Jeffrey Cole. I've been working with this data for nearly two years now and we have all been waiting for the long promised longitudinal analysis, which, as it turned out, wasn't possible due to some bad initial data-collection decisions. No matter though, it seems the project has no problems making causal claims despite having only cross-sectional data and explicitly comparing Internet users and non-users!
"Use of the Internet is reducing television viewing around the world while having little impact on positive aspects of social life," said Jeffrey Cole, director of the UCLA Center for Communication Policy, the California university that organized the project.
What??? While the claim itself might not be wrong, Jeffrey Cole, I dare say, has no data that would allow him to make this claim! Cross-sectional analyses produce correlations but compare essentially different populations (of users and non-users). Yes it is possible to use statistical controls to attempt to control for differences between these populations, but that is both error-prone and invariably incomplete. It is simply impossible to measure ALL differences between these populations. The very decision of whether to use the Internet or not differentiates the two populations, that also happen to differ on many other aspects. Yes there maybe a strong statistical association between TV watching and Internet use, yet that is no basis for a causal claim! Such blunders are infuriating, especially when presented in mass media, where readers lack statistical and methodological training to realize these mistakes and not to take these results on face value! This type of data does not allow us to say that Internet use in effect causes declines in TV watching. There maybe a host of other reasons for why these populations have different patterns of both TV watching (insert sociability, depression, any other variable of interest here) and Internet use and none of those are considered.
On another point - simple comparisons between users and non-users are simply insensitive. People have extremely different patterns of Internet use both in breadth of its applications and in its intensity. Simply lumping all Internet users together may lead to erroneous conclusions. For example, what if people who use the Internet a lot but only for entertainmnet, actually watch TV more, but those that use it for communication or news, watch TV less? What if people who use the Internet only a little substitute that activity for TV watching simply because they are just not big media users and they get their "fill" from Internet now. Wouldn't that change the picture? yet the project's analyses would never be able to detect these distinctions yet their conclusions certainly "sound" strong in the article!
source: Many2Many
No comments:
Post a Comment