Altmetric supports the use of its data in the development of journal-level indicators and metrics that adhere to “responsible metrics” practices, especially when such metrics are used as part of a multi-indicator or “basket of metrics” approach to understand journal-level influence.
In general, Altmetric supports the development of journal-level indicators and metrics that:
Account for known skew using coverage percentiles (e.g. “% articles with an Attention Score in a journal”), geometric means, or medians (technically allowed, but discouraged).
Address biases related to subject area, location, and other confounding variables, for example by using subject-area normalization.
Look beyond the boundaries of specific journals towards network analysis, communications theory, and topic-based analysis.
Are transparent and otherwise in line “responsible metrics” practices described by the Leiden Manifesto.
We encourage that researchers think carefully about using the correct denominator in developing their journal-level metrics. Altmetric data typically does not include the entirety of all articles published by a journal; instead, we index only those articles that have been mentioned in a source we track (with a few exceptions). It’s better to use an independent source (e.g. Dimensions, Scopus, Web of Science) that provides reliable and auditable data to determine a reputable denominator.
Analysts should avoid using basic calculations like sums or averages based upon the Altmetric Attention Score or source-specific counts (e.g. number of tweets). These metrics do not account for the relatively large amount of outliers that exist in any given sample.