We begin by examining the variation in the total number of flu cases reported each season for New York and Missouri. JMP’s Graph Builder is a flexible platform that allows many different types of graphs to be created.

Open flu_NY_MO.jmp and invoke Graph > Graph Builder from the JMP Menu. Drag the variable, Total Influenza to the Y drop zone, Season to the X drop zone, and State to the Group Y drop zone. Choose the Bar chart icon in the control panel and select Sum from the Summary Statistic drop-down menu. The completed Graph Builder dialog is shown in Figure 8.1 Graph Builder Dialog to Create Bar Charts of Total Influenza Cases by State.

This visualization shows the total number of flu cases reported in each season by state. Notice that the Y-axis is automatically scaled to the same range for both bar charts. This facilitates comparison between the two states.

The flu activity data is given weekly for each of the seven seasons. In the bar charts of Figure 8.1 Graph Builder Dialog to Create Bar Charts of Total Influenza Cases by State the Summary Statistic chosen was Sum. This aggregated the total cases for a season across all weeks. A more detailed comparison of the flu seasons can be made by looking at the distributions of weekly flu activity. Histograms are commonly used to display data distributions. With the Graph Builder dialog set as in Figure 8.1 Graph Builder Dialog to Create Bar Charts of Total Influenza Cases by State, choose the Histogram icon from the control panel. Figure 8.2 Histograms of Weekly Influenza Cases by Season and State shows the resulting graph.

This matrix of histograms exemplifies the data visualization technique of “small multiples” popularized by Edward Tufte. This method of multivariate data display uses similar graphs with the same axis scales sequenced over one or two other variables. The advantage of small multiples is that the observer can focus on changes in the data rather than changes in the graphical elements. Figure 8.2 Histograms of Weekly Influenza Cases by Season and State is an example of a small multiple sequenced over Season and State.

Unfortunately, the skewness of the weekly flu activity coupled with the display of 14 histograms makes it difficult to discern differences across either seasons or states. Box plots are an alternative way to visualize data distributions. In a box plot, a data distribution is summarized by a box whose ends lie at the 25th and 75th percentiles. The line inside the box shows the median. The “whiskers” extend (1.5*Interquartile range) beyond the ends of the box. Outliers beyond the whiskers are shown as dots. While box plots have less detail about the distributional shape than histograms, they are effective when comparing a relatively large number of groups in a small space. To create a matrix of box plots, select the box plot icon from the Graph Builder control panel. The result is shown in Figure 8.3 Box Plots of Weekly Influenza Cases by Season and State.

This visualization is more effective for comparing the weekly distributions of flu activity than the histograms shown in Figure 8.2 Histograms of Weekly Influenza Cases by Season and State. The choice of graph type (e.g., box plot, histogram) that is most effective will depend on the data distributions and the number of small multiples to be displayed. Graph Builder’s control panel makes it easy to evaluate different graphs.

Data visualizations can be improved by judiciously annotating graphs with problem domain information. During the seven seasons being analyzed, the H1N1 pandemic occurred in the 2009-2010 flu season and in 2012 the FDA approved the use of cell-based technology for manufacturing flu vaccine. This new technology allows vaccines to be produced more rapidly than the traditional egg-based manufacturing process.

The JMP Annotate tool allows text boxes to be added to graphs. Figure 8.4 Total Influenza Cases Bar Charts with Annotations Added shows the bar charts for total influenza cases with the addition of two text boxes.

When annotating graphs, place text boxes on charts so that they do not obscure the graphical elements. Avoid excessive annotation, which will detract from the effectiveness of the graph. Annotations provide additional information that improves the interpretation of the data in the context of the problem domain.

Figure 8.4 Total Influenza Cases Bar Charts with Annotations Added allows us to easily compare the Missouri and New York flu activity by season. Both states experienced the highest level of flu activity during the 2014-2015 season. New York’s flu activity remained relatively high from 2012-2013 to 2015-2016 while Missouri only experienced the high level in 2014-2015. Both states saw relatively low flu activity during the 2009 pandemic.

During a flu season, activity can be monitored on a weekly basis. Again, Graph Builder offers a flexible platform for data visualization. In order to look at the 2015-2016 season for changes week-to-week, open the 2015-2016 flu season JMP file, created in the Data Management section. This contains weekly data for both New York and Missouri for the 2015-2016 flu season.

Select Graph Builder from the JMP menu. Drag Total Influenza to the Y drop zone. Drag Week Ending Date to the X drop zone. Notice that the dates are placed in chronological order along the axis. In the control panel select the line icon. Drag State to both the Overlay and Color drop zones. The completed Graph Builder dialog is shown in Figure 8.5 Graph Builder Dialog to Create Line Graphs for Weekly Flu Activity for the 2015-2016 Season.

Each state’s weekly flu activity is displayed as a separate line graph on the same X and Y axes. Place the cursor anywhere over one of the lines and right click. Click Add > Points to add markers to the line graphs as shown in Figure 8.6 Graph Builder Dialog to Add Markers.

Click Done to obtain the final graph shown in Figure 8.7 Line Graphs of Weekly Influenza Cases by State.

Different markers enable the lines corresponding to each state to be distinguished when the graph is reproduced in black-and-white.

When there are multiple line graphs displayed on a single set of axes, the lines may be difficult to distinguish. In these cases, creating a small multiple display may be preferable. This can be accomplished in Graph Builder by dragging State from the Overlay drop zone to the Y groups drop zone. The result is shown in Figure 8.8 Line (Smooth) Graph for New York and Missouri for the Weeks of the 2015-2016 Flu Season, with Overlay by Color.

Both Figure 8.7 Line Graphs of Weekly Influenza Cases by State and Figure 8.8 Line (Smooth) Graph for New York and Missouri for the Weeks of the 2015-2016 Flu Season, with Overlay by Color show that at the peak of the flu season, New York has more cases reported than Missouri. This is to be expected as New York has a larger population than Missouri. For both states flu peaks at roughly the same time (early March).

Finally, we create a small multiple display that shows the pattern of flu activity for the two states for flu types A and B. In Graph Builder drag Influenza A to the Y drop zone then drag Influenza B to the Y drop zone. This will create two Y axes, one for each flu type. Drag State to the X group drop zone and Week Ending Date into the X drop zone. Select the line icon from the control panel and add markers as described above. The resulting graph is shown in Figure 8.9 Small Multiple Display of Weekly Flu Activity by State and Flu Type for the 2015-2016 Season.

In this display, we see the differences between the two strains of flu with the peaks occurring at the same time but lasting longer for Influenza B. This pattern was consistent between both states.