Friday, January 6, 2012

Credibility Intervals for Medical Journals

A paper by Darren Greenwood in BMC Medical Research Methodology estimates confidence intervals for impact factors based on observed fluctuations in the impact factor over time. He comes up with these intervals for journals in research and experimental medicine in 2005:

As you might expect, the credibility intervals * are wider for lower ranked journals. These estimates do not use actual citation data to individual articles, as I did, and Greenwood notes in the article that using such data would further widen the credibility intervals.

It is pretty clear that confidence intervals for impact factors for low ranked journals are high. The providers of these statistics should really provide a measure of precision. I'm surprised that this doesn't seem to be an extensively researched area in bibliometrics.

* Bayesian equivalent to confidence intervals.

No comments:

Post a Comment