With the usual flurry of Impact Factor announcements due to start any day now, it’s a good time to remember thatIt also shows the graph below of the citation distribution for the journal. Note how incredibly broad and asymmetrical the distribution is. I found this interesting because several years ago I wondered what the error bar was on Impact factors, which are often reported to several decimal places.it is the papers, not the journals they´re published in, that make the impact.

The editorial points out that the distribution is probably broader for PLOS ONE because it publishes articles from a diversity of fields. Hence, I would still like to see distributions from other journals.

Nature Materials published their citation distribution a couple years ago. It was Poisson distributed, so I downloaded the data and fit it to a decaying exponential, which gave me a mean of 34 and an error of 2. I think impact factors should be thus reported to the nearest integer with the error reported as well.

ReplyDeleteThanks for the very helpful comment. I found the relevant editorial including the distribution curve

Deletehttp://www.nature.com/nmat/journal/v10/n9/full/nmat3114.html

Here is an extract:

"in the skewed distribution of a journal's received citations, which are known to approximately follow a Pareto probability distribution — also known in bibliometrics as Bradford's law3. The Pareto distribution is a power-law function characterized by a long tail — in this context a long tail of highly cited papers "

I recommend it to readers.