Positional analysis: from boring tables to sweet visuals
March 28, 2018
, , , , , , ,

At Science-Metrix we are obviously very focused on data—sweet, sweet data! We are also very aware that bibliometric data or pages and pages of analysis can be overwhelming and that more intuitive data presentation can help our clients to better understand their study results, which in turn helps them to take action on the findings we return to them. One graphic presentation we find particularly helpful is the positional analysis chart. Positional analysis is a way to visually depict two, three or even more indicators for a given set of entities instead of using a standard (and boring) table. Here’s how it works.

First, let’s pretend you have a table containing a list of countries doing research in nanotechnology, and each country possesses three attributes: the sum of its nanotech scientific articles, their citation impact (using the average of relative citations, or ARC) and a nanotech specialization index. I know you’ve seen a thousand tables just like this, all of which seem to be completely identical, featuring numerous lines, with many numbers per line. I can hear you yawning already!

positional analysis, data visualization, bibliometrics, Science-Metrix, strategic planning tools

So, how can we synthesize all this information in a simple visual? The backbone of a positional analysis is a 2D Cartesian coordinate system, where each axis represents an index or an indicator. By positioning a point on the coordinate system, you automatically assign two values to this point. Going back to our country table described above, let’s set the specialization index as the x-axis (horizontal) and the citation impact as the y-axis (vertical), with the world average of each index set at the origin (where the major axes meet).

Here is a simple example of how it might look (note, I’ve used fictitious data):

positional analysis, data visualization, bibliometrics, Science-Metrix, strategic planning tools

By positioning each country according to its values of citation impact and specialization index, it’s easy to identify which countries perform better than the world average on each of the indices. Countries are more impactful than the global average if they’re above the major horizontal axis, less impactful if they are below it. They are also more specialized than the global average if they’re to the right of the major vertical axis, less specialized than average if they are to the left. This makes it easy to rank countries at a glance with respect to each other: who’s above or below whom, and to the left or to the right of whom.

The observant among you will have noticed that there is another indicator represented in the chart above: it’s the number of articles produced by each country and it’s represented by the size (or area) of the country’s circle. In looking at this, we should note that size of production and specialization are separate but linked properties. A large bubble on the left side of the chart shows that the country’s production is large (demonstrated by the size of the bubble) but that the field is still not one of specialization for it (demonstrated by its position to the left of the centre line). Such a situation would indicate that the country is producing many papers in a field with a large publication output on the world stage—a relatively smaller drip in a huge pool. Conversely, a small bubble on the right side of the chart would denote a subfield in which the country is not producing many papers (demonstrated by the small bubble), but that this subfield is small at the world level, and therefore the country’s small output in this subfield is still above the global norm (demonstrated by the position of the bubble to the right of the centre line).

Reading the figure above, we can easily see that the largest number of nanotechnology articles are published by France (FR), the United States (US), Canada (CA) and Germany (DE). Among those four countries, France obtained the higher citation impact, while Germany obtained the lowest, being cited even less than the world average. (Recall that the data used are fictitious; I don’t mean to offend any country here!) In terms of specialization index (share of a country’s articles in nanotech over the share of the world’s articles in nanotech), our fictitious Canada and Germany are the most specialized within the four largest publishing countries mentioned above.

We could also consider a fourth indicator to be presented on the very same chart, one based on the colour of the circle. For example, shades of green and red could be used to denote the growth of papers in nanotech over a given period. The example above uses colour for aesthetic purposes only, but colour is one way to synthesize even more data within a single visual. We could even make separate positional analyses as annual snapshots, and then animate them as a GIF to show evolution over time, integrating a fifth dimension into our visual.

Positional analysis charts help to communicate a lot of information in very little time. They’re also excellent for highlighting the relationships between different attributes, and they are a very useful aid in strategic planning, through identifying the strengths and weaknesses of a given entity. I’ve demonstrated one example here using countries and some bibliometric indicators, but in fact a positional analysis can be used to represent all kinds of entities and all kinds of data, whether bibliometric or not.

If you’d like to see these charts applied to some real-life data, head over to our Twitter feed. Throughout 2018, we’ll be rolling out a Twitter series of positional analyses presenting the research performance of countries (USA, China) or groups of countries (EU, South America) across the fields of science.


Note: All views expressed are those of the individual author and are not necessarily those of Science-Metrix or 1science.


About the author

Bastien St-Louis Lalonde

Bastien St-Louis is a Research Analyst at Science-Metrix. He holds a BSc in Physics from McGill University and an MSc in Physical Engineering from Polytechnique Montréal. He uses his strong theoretical background to compute and develop bibliometric indicators at Science-Metrix, where he has worked since 2013. In that time, Bastien has been involved in more than 45 bibliometric studies, conducted for a wide range of clients including Canadian and international government agencies and departments, and international NGOs such as UNESCO and CGIAR.

Related items

/ You may check this items as well

Rationalizing the extremes: introducing the citation distribution index

The distribution of citations among the scientific...

Read more

1findr: discovery for the world of research

As of last week, 1science is offering public acces...

Read more

Mapping science: a guide to our Twitter series

Over the course of 2018, we’ll be publishing a s...

Read more

There are 7 comments

  • Really lovely! You guys are the best!

  • Sandy says:

    I am new to this blog – but what I am seeing is great. What software was used to do the figure?

    • Brooke Struck says:

      It’s actually done in Excel using macros and some manual beautification. For all the quantitative heavy lifting that goes into conditioning, cleaning and computing the underlying data, creating a good visual remains in large measure an art!

      Any applications that you have in mind for this type of visual? Would love to hear some examples in the comments!

  • Graydon Snider says:

    It would be a nice, data permitting, to add a time slider to observe the chart evolution, e.g. Combined with a menu option for retrieving the raw data as a csv/xlsx/etc, it eliminates need to choose between tables and graphs.

  • Robert Chen says:

    Very interesting post!

    I’ve been seeing more and more of these to present information in non-academic circles such as for hockey statistics. However, one of the accompanying criticisms I have seen with these type of graphs is that they can grow into unintelligible monsters that have too many “axes” to be intuitive. Do you have any suggestions on whether there is a limit to how much information a single positional analysis graph should contain?

    • Bastien St-Louis Lalonde says:

      Dear Robert,
      Thank you for your kind words.
      You are right; trying to pack too much information would bring us back to square one where the visual would not be intuitive anymore!
      We usually represent only three dimensions on such visuals. Two represented by the x and y axis and a third one with the size of the disc. Sometime we go to four, adding a disc color to represent a fourth dimension (ideally with a legend). I would advise against more than four dimension as it would then become more confusing than anything else!