Measurement of research activity still remains a controversial question. The use of the impact factor from the Institute for
Scientific Information (ISI) is quite widespread nowadays to carry out evaluations of all kinds; however, the calculation
formula employed by ISI in order to construct its impact factors biases the results in favour of knowledge fields which are
better represented in the sample, cite more in average and whose citations are concentrated in the early years of the articles.
In the present work, we put forward a theoretical proposal regarding how aggregated normalization should be carried out with
these biases, which allows comparing scientific production between fields, institutions and/or authors in a neutral manner.
The technical complexity of such work, together with data limitations, lead us to propose some adjustments on the impact factor
proposed by ISI which — although they do not completely solve the problem — reduce it and allow glimpsing the path towards
more neutral evaluations. The proposal is empirically applied to three analysis levels: single journals, knowledge fields
and the set of journals from the Journal Citation Report.