From Visualization Sp06
This visualization is designed to show the predicted spread of a flu pandemic if bird flu were to begin spreading between humans. It shows predictions for the first two months after the flu arrives in the United States, if it were not slowed by vaccines or antiviral drugs. That is, it shows what would happen if we failed to do anything about the threat of a pandemic.
The audience for the visualization is readers of Scientific American. Readers are almost exclusively college educated, 73% male. Most readers have careers in business, not directly related to science. The magazine is designed "to inform people who are in leadership roles". As such, the visualizations are (or should be) designed to give readers the information they need to stay on top of scientific issues that may affect their businesses.
The data set for the visualization includes the predicted number of symptomatic flu cases per 1,000 people for each county in the United States, with predictions for 15, 30, 45 and 60 days after the beginning of an outbreak (~12,000 values). The data set also implicitly includes the latitude, longitude and shape of each county and state. This data is translated into a small multiple visualization. Each of the four subcomponents uses the X and Y axes to encode latitude and longitude and uses hue to encode the number of flu cases in each county (all quantitative variables). Time is encoded by encoded by the position of each map, although it does not map directly onto the X or Y axis (thus the need for explicit time labels on each map).
This visualization does a good job of making its high level point, which is that a flu pandemic would affect many United States residents (especially in densly populated areas) within a couple months of an outbreak. While it makes use of color hue to encode the quantitative variable number of cases per 1,000 residents, it does so because of the cultural convention associating green with "good" and red with "bad". Even without reading the associated text, a quick glance at the first and last map makes it clear how quickly the pandemic would spread.
In designing this visualization, I might consider using a color axis including only red and green hues (as in Map of the Market) because the blue and yellow hues used in the visualization lack the same cultural meanings as red and green. However, chosen spectrum was likely used to avoid the implied "zero point" of an axis including only red and green. This choice makes it difficult to compare values across different counties.
Additionally, because the visualization encodes information at the county level, it suffers from the fact that counties in the western United States are bigger than counties in the eastern United States, resulting in a different resolution of data in the two regions of the country. (This is particularly apparent in the "Day 45" map.)
The table above lists the integration values for each of the rooms in the floor plans shown below (Figure 1). The integration value of a room is a measure used in the Space Syntax analysis of architectural spaces that measures the "connectedness" of that room to all other rooms in a building. For example, a highly connected and centralized room (e.g. room 4 of house gamma) would have a higher integration value than a room in the corner connected to only one other room (e.g. room 4 of house delta). Integration values are used in the social analysis of space -- often, the integration values of different rooms will be compared to the roles of people using different areas of a building (men vs. women, adults vs. children, royalty vs. servants, etc.).
This example is used in the introduction to Decoding Homes and Houses to explain the intuitive meaning of integration values for a diverse set of sample homes (alpha, beta, gamma, and delta). In addition to the floor plans and the table of integration values, the example also includes justified graphs of each floorplan (Figure 2), which encode the connectivity between pairs of rooms in the home.
The data for the table includes includes the quantitative variables of integration values for each room in each of the four houses, both excluding and including "the outside" as a separate space (columns 2 and 3, respectively). It also includes the mean integration value for each house (with and without the outside) and the difference factor (DF), a measure of the variation present among the rooms in each house (84 values). These values are arranged in the table, with four spans of three columns each, one for each house. Within each span, there is one vertical column with room name, one with the integration value without the "outside", and one with the integration value with the "outside".
The justified graphs of the floorplans use distance between pairs of nodes to encode the length of the path from the room represented by the root node to each other room in the house. The nominal variable of a doorway between pairs of rooms is represented by connections between nodes in the graph.
Aside from horrible table design e.g., little round boxes around each table, and summary variables, (Mean and DF) that are not differentiated from the rest of the table, the table doesn't accomplish the goal of helping to explain the intuitive meaning of "integration value" in the context of the sample floorplans. First, the rows are not sorted by integration value, making it difficult to compare the relative values for each of the different rooms. Second, the table doesn't encode any of the relationships between rooms, making it difficult to compare the value to the position of each room within the house. Because the table is shown on a different page from both the floor plans and the justified graphs, it is even more difficult to compare values to their actual location in the physical layout of each home.
In viewing this visualization, the goal of the reader is to see the relationship between integration values within each house and across the different floorplans, specifically pointing out the differences that arrise across floor plans that look similar at first glance. This could be better encoded by using value to shade each room in the floor plans and node in the graphs proportional to its integration value, allowing the reader to assess at a glance the differences between the different rooms. A table might be useful for allowing more specific lookups and comparisons, but only in combination with other graphics that make it easy to grasp the high level picture.