Viewer's Choice Presentation

Back to Homepage

1. Introduction

This visualization depicts traffic accident data in the city of Medellín, Colombia between 2014 and 2018. The visualization was built by Jorge Iván Pérez Rave of the Colombian consulting firm IDINNOV, seemingly as a demonstration of their data visualization capabilities.

2. Data

This data is sourced from the Colombian national government (datos.gov.co) and Medellín city government (medellin.gov.co), specifically the GEO Medellín Project. The visualization does an exemplary job of citing sources not only for the data used, but also for code and libraries used for the construction of the site. There is little information, however, of how this data was processed to produce the graphs shown.

3. Visualization Critique

"GRAVEDAD DE LOS ACCIDENTES VIALES" - GRAVITY OF ROAD ACCIDENTS

This is a bar graph showing the relative proportions of traffic accidents, divided into the categories "Herido" - Injuries, "Solo Daños" - Only damage (i.e. no injuries), and "Muerto" - Deaths.

Good:

Bad:

"SEGMENTACIÓN DE ACCIDENTES VIALES" - SEGMENTATION OF ROAD ACCIDENTS

Several graphs are concealed by a dropdown menu here, labeled "Segmentos" - Segments.

The first segment is "Hours". It took me a long time to understand what was happening in this graph. This is a horizontally oriented bar graph in which the vertical axis represents hour of the day - bars representing the percentage of accidents which take place at that hour of the morning extend to the left, and for the evening extend to the right, from a central 0 axis.

Good:

Bad:

One of the more interesting graphs in this section is a bar graph of accident frequency by type of street or intersection. It appears from this data that the vast majority of accidents occur in locations of the category "Tramo de via" - tramway or train tracks.

Good:

Bad:

"MAPA DE ACCIDENTALIDAD" - ACCIDENT MAP

The locations of fatal accidents are marked on a map. From the map, it appears that most fatal accidents take place on major streets.

Good:

Bad:

"CALLES Y CARRERAS QUE DESCRIBEN LA ACCIDENTALIDAD"

I find this one of the more interesting graphs in the project. The streets of Medellín are broken up into Calles, which run East-West, and Carreras, which run North-South. This graph treats each Calle and Carrera number as a coordinate, and in so doing normalizes the accident locations into a rectangular grid. This allows us to see quite clearly which streets are the most dangerous, as these appear as straight lines on the graph. However, there are shortcomings of implementation which keep this from living up to its promise.

Good:

Bad:

"EVOLUCIÓN DE LA ACCIDENTALIDAD" - EVOLUTION OF ACCIDENTS

This graph shows the total number of accidents by month for the years between 2014 and 2018.

Good:

Bad:

MÉTRICAS SOBRE COMUNAS - Metrics By Commune

This is a set of graphs which break up the city of Medellín into its constituent communes, or neighborhoods. The metric being graphed is chosen with a dropdown menu. To the non-Spanish speaker, these labels range from cryptic to completely indecipherable, a valuable reminder of the accessibility cost of nonstandard abbreviation, but fortunately a fuller description of most graph are available after selecting them.

Good:

Bad:

"RELACIONES BIVARIADAS DE MÉTRICAS SOBRE COMUNAS" - Bivariate Relations of Metrics By Commune

This graph addresses many of the failings of the previous graph by allowing the user to compare any two metrics of their choice. In the upper left is a histogram for variable 1, in the bottom right a histogram for variable 2, and in the bottom left a scatter graph relating the two, with a centroid and some sort of trend line marked. The large number in the upper right is the Kendall rank correlation coefficient, a measurement of assotiation between the chosen variables.

In this picture, I have chosen to graph total accidents against quality of public transport connections. This shows a moderate correlation, which may be explained by the tendency of areas which are well-served by public transportation to also have heavy traffic.

Good:

Bad:

"PATRONES DE AGRUPACIÓN ENTRE COMUNAS" - Patterns of Grouping Between Communes

This is another interesting graph, as it relates not to absolute values or proportions of metrics, but to how similar any two communes are to one another.

Good:

Bad:

FACTORES LATENTES ASOCIADOS CON EVENTOS DE ACCIDENTALIDAD USANDO REGRESIÓN MÚLTIPLE y ACP. DATOS 2016 - Latent Factors Associated With Accident Events Using Multiple Regression and Acp. Data 2016

This section is undeserving of the name of visualization. It is a table describing the multiple regression techniques being used, accompanied by a rudimentary line graph. It is very unclear to a lay audience (like myself) what is happening here.

JORNADA Y HORARIOS ASOCIADOS CON EVENTOS DE ACCIDENTALIDAD BAJO ENFOQUE PARETO. DATOS 2014 - 2018_1 (hasta mayo 2018) - Day and Schedules Associated with Accident Events

This section is the only one to break down each category of accident (fatal, injury, damage only) into subcategories, which are very interesting. For example, we can see from this graph that most accidents in La Candelari are caused by "Atropello" (Being run over), followed by "Choque"(Shock), and only a very small proportion being "Otros"(Other). This would seem to imply that most victims of traffic accident are not drivers, but bystanders.

Good:

Bad:

4. Conclusion

I would not hire this firm. Despite some innovative presentation, details are consistently neglected, and data is presented without meaningful conclusions being offered.