## Thursday, August 25, 2011

### How to use the h-index

A few have asked me how to use the h-index in light of what I showed earlier. In many cases, the h-index is used for promotions, for example. For assessments, I would recommend not just looking at a person's h-index, but instead examining the residual h-index and finding good comparables.

Quantity and quality: residual H-index
The H-index is supposed to represent scientific productivity beyond just the number of publications. Yet, 90% of the h-index is the number of publications and the time a person has been publishing. It’s actually the residuals that are the key here. Two individuals with the same number of pubs and years publishing could differ in their h-index, if one is cited more. Assuming the number of citations correlates with publication quality, then the person with the greater residual h-index would have a greater impact.

There are always caveats to this, but it’s clear that for the purposes of assessment, one should examine the number of publications and the time a person has been publishing as well as the residual H-index from scientists in the discipline. This is probably the best metric of impact beyond number of papers.

Find comparables.
One of the benefits of the approach is to be able to find comparables. Just like in real estate, appraisals are used to determine the potential market value of a house and are anchored with the sale value of comparable houses. Just like researchers, no two houses are exactly alike (except in some uninteresting subdivisions), but they can be compared.

My approach to finding comparables is to generate the relationship between the H-index and the number of publications and the number of years publishing. Then, determine the next closest people in the space defined by the actual and predicted H-index. For example, calculating a Euclidean distance between my scores (H-index = 22, predicted = 20.4), the next closest person to me is a friend of mine, actually: 80 pubs in 12 years, H-index of 24, predicted 23.4. Euclidean distance = 3.6.

The person furthest from me? My advisor, Terry Chapin. 321 publications—probably more with some misspellings. H-index of 84. Predicted H-index of 80. Distance of 86.

Here’s a graph of distance from my scores as a function of H-index for reference.

Even objective metrics have subjective assumptions. Still, there are important lessons to be learned from quantifying scientific productivity. Might as well do it as well as possible.