Working with data from data.ny.gov and using Tableau this visualization explores leading causes of death from 2007 to 2011 focusing on the top causes and comparing those by sex, by ethnicity, and looking at biggest trends over time.
As causes of death are so numerous, the separate visualizations are limited to top causes. Heart disease is by far the most prevalent cause of death in both men and women, however between 2007 and 2011, more women died of heart disease. Over that same time period the disease shows a drop in number of deaths, whereas deaths by cancer increased. Significantly more black people died from diabetes than any other ethnicity according to the data.
The top visualization gives an overall view of causes of death over the given time period. As this is a part-to-part analysis and the main goal is to convey a general sense of the quantitative data but easily locate the categorical data, a circle graph was used. As the exact numbers are not necessary to communicate the message here and the focus is on the nominal scale, this seemed appropriate. However, for the other graphs in which comparison of quantitative data is more important to the communication goal a more precise correlation necessitates using two of the most powerful attributes of visual perception for encoding quantitative values, line length and 2-D position.
The bar graph comparing causes of death by sex is a simple nominal comparison relationship. This graph might be more meaningful if arranged in a ranking relationship with male and female next to one another for a more accurate visual comparison. It might also be more interesting if the values were a bit more exact as certain comparisons look almost equal on this graph but might show more differentiation if looked at in greater detail. It could also be revealing to explore the relationships here in a part-to-whole graph and get a better sense of the percentages these represent in relation to the total number of deaths.
The bar chart exploring deaths by ethnicity might be more effective if the values had been normalized in some way as there were significantly more white people in this population. That being said, the fact that the diabetes section is entirely green—indicating non-hispanic black—is significant. It would also be interesting to take a separate look at diabetes in this case as it is a disease that is very affected by socio-economic factors.
The line graph was used to chart diseases over time, emphasizing the shape of the data as it moves from value to value. The graph shows a significant drop in heart disease. However, the graph is displayed in such a tight vertical space, that this might create the perception that this trend is more significant that it actually is. This chart would be optimized if given more horizontal space. In addition, as it provides an overview of the data before delving into more specifics such as the sex and ethnicity graphs, it might be more effective more prominently displayed to give a better sense of hierarchy on the dashboard. A clearer arrangement of the data, with appropriate placement of information based on importance and desired viewing sequence would benefit the dashboard.
Future directions would include exploring a better use of color. There are too many colors involved in this visualization which detracts from its clarity. For instance, the graph focusing on causes of death by sex might not even need any color coding but would work if simply labeled with the diseases and featuring a a varying gray scale to differentiate between male and female. Currently, with labels and color coding, the viewer is forced to process excess levels of data. The extra legend on the top right also takes up valuable real estate.