Mathew Kiang (.com)Yes, with one "t".

· 5 min read

A visual tour of my publications

A line chart titled Citation history over time. One grey line per article, with year from 2010 to 2016 on the x-axis and cumulative citations from 0 to 25 on the y-axis. Each line begins at the article's publication year, so lines start at every year across the chart and the newest cluster at 2015 and 2016 near zero. Four lines reach 20 or above by 2016; most end between 1 and 14. A very pale straight line runs across the middle as an overall fit. With every line the same colour, no group is distinguishable.

I recently came across this paper by Michal Brzezinski about (the lack of) power laws in citation distributions. It made me a little curious about the citations of my own articles so I threw together a little script using James Keirstead’s Scholar package for R. In the plot above, every line represents a single article with time on the x-axis and (cumulative) number of citations on the y-axis.

It’s not super informative, so we can break it down a few ways to graphically explore the data.

The easiest thing to do is to simply color-code each article by the type of research it involves. My research spans four overlapping phases: (1) My work with CHIBPS studying the intersection of HIV, drug use, and behavior among young men who have sex with men in New York City; (2) my two years as a research assistant in Health Policy and Management studying surgical safety checklists, implementation of health initiatives, and patient-centric care; (3) the start of my doctoral program and studying socioeconomic and racial disparities in health; and (4) my more recent (and concurrent with 3) work on digital phenotyping with the Onnela Lab.

Once each article is color-coded, we can see my research transition over time below. The grey line represents the best fit (linear regression) line across all articles and groups.

The same chart as above — Citation history over time, cumulative citations 0 to 25 against year 2010 to 2016 — but each article's line is now coloured by one of four article types: Digital Phenotyping in green, Health Inequalities in orange, Health Policy and Management in purple, and MSM / HIV / Drugs in pink. A thick grey straight line runs across the middle as the overall linear fit. The pink lines start earliest and include all four of the highest-reaching articles; purple starts around 2011; orange lines all begin in 2014 or later and climb steeply from near zero; the single green line begins in 2015 and reaches 3.

What if we were interested in the actual slope of each group? That is, do some areas appear to get cited more quickly than other research areas? Plotting the same data as above, but fitting the regression line for each group, we can see the slope for all groups are relatively similar except for Health Policy and Management, which appears to have picked up new citations a little slower than the others.

A chart titled Citation history, fitted by type, over time. Cumulative citations 0 to 25 on the y-axis against year 2010 to 2016. The individual article lines are faded to pale grey and one thick fitted straight line is drawn per group, for four article types: Digital Phenotyping in green, Health Inequalities in orange, Health Policy and Management in purple, and MSM / HIV / Drugs in pink. Pink is the steepest, running from 0 in 2011 to about 16 by 2016. Orange and green are similarly steep but start much later, in 2014 and 2015. Purple is by far the flattest, from about 1.5 in 2011 to only 7.5 in 2016.

We can show this more clearly by shifting all articles to 0 and instead plotting the age of each article. Visualizing it this way shows my health inequalities articles picking up citations the fastest and confirms HPM articles (at least my HPM articles) picking them up more slowly.

A chart titled Citation history, fitted by type, over article's age. Citations 0 to 25 on the y-axis against years from publication, 0 to 5, so every article now starts at the origin. Pale grey lines are individual articles and one thick fitted line is drawn per group, for four article types: Digital Phenotyping in green, Health Inequalities in orange, Health Policy and Management in purple, and MSM / HIV / Drugs in pink. Orange is the steepest, reaching about 12 citations by age 2 — the age its newest articles reach. Pink is next, about 20 by age 5. Purple is clearly the shallowest of the three that extend that far, about 14 by age 5. Green, the newest area, only reaches age 1.

Now, suppose we want to know what the composition of my citations in regards to article type over time. We can plot a stacked bar chart below with each bar representing a year and the y-axis representing number of citations. Again, colors represent article type. Unsurprisingly, we see my older work taking up a smaller proportion of the total number of citations as time progresses.

A stacked bar chart titled Citations per year by article type, one bar per year from 2010 to 2016, citations 0 to about 64 on the y-axis, stacked by four article types: Digital Phenotyping in green, Health Inequalities in orange, Health Policy and Management in purple, and MSM / HIV / Drugs in pink. The total grows from 1 citation in 2010 to about 64 in 2016. Pink is the whole bar in 2010 and 2012 and most of it through 2014, then shrinks to roughly a quarter of the 2016 bar. Orange first appears in 2014 and is the largest single segment by 2016. Purple appears from 2011 onward as a smaller band. Green appears only in 2016, as the small segment at the bottom.

No real take home message here. Except maybe if you’re looking for ways of increasing your citation count in the field of public health, health inequalities seems to be a good way to go — which is fortunate because you won’t be making any money. Also, there seems to be a bug in Google Scholar. It shows my total citations as being over 200, yet when counting the citations for each individual article, I’m still well below that. Code here.

Acknowledgement: Helpful comments by Monica Alexander on the first plot inspired the rest of this post.