I conducted the research described here in 2006, utilizing data from the annual census of U.S. public libraries for 1999 through 2004. These data were originally made available by the National Center for Educational Statistics (NCES), and are currently available at the Institute for Museum and Library Services (IMLS) website.
A method common to many studies of U.S. library statistics is to separate libraries into groups based on some basic demographic variable, usually the population of the communities which libraries serve, referred to by NCES as "service areas". Thus, this study uses the following groupings (originally defined by NCES):
| LIBRARY SERVICE AREA POPULATION RANGE |
ABBREVIATION USED IN STUDY CHARTS |
| Less than 1000 | <1K |
| 1,000 to 2,499 | 1K |
| 2,500 to 4,999 | 2.5K |
| 5,000 to 9,999 | 5K |
| 10,000 to 24,999 | 10K |
| 25,000 to 49,999 | 25K |
| 50,000 to 99,999 | 50K |
| 100,000 to 249,999 | 100K |
| 250,000 to 500,000 | 250K |
| More than 500,000 | >500K |
The graphic directly below is a sample chart showing ten geometric shapes, each aligned directly above the abbreviation for the population group it refers to. (The shapes and symbols used in boxplot-style charts are described in detail in the Guide to Interpreting BoxPlots.) Note how the shapes (boxplots) towards the right of the chart are quite flattened. This is because the distributions of library statistical measures (like circulation, volumes held, expenditures, and so on) vary significantly depending on service area population. For this reason, statistical data items listed here are depicted in charts that come in sets of three (3). The charts are therefore labeled as A, B, and C in charts 13A through 37C.*
Charts labeled A present high level views ("overviews") of distributions of a given statistical item (called a "data element") for the ten library groups. For example, chart 13A shows ten separate boxplots, aligned left to right, for the service area population data element. Note that the left axis in the chart ranges from zero to four million. Such an extended scale causes the boxplot-style symbols on the left side of the chart to be compressed nearly into flat, horizontal lines.
To compensate for this problem, two other charts in the set present a more readable ("closer") view of the same data. See charts 13B and 13C. (After clicking on either link, scroll down to see B and C charts.) Note in these charts how the boxplot shapes are no longer flattened. This is because these two charts allow us, in a sense, to "zoom in" on the data to take a closer look.
In A charts, the range indicated by the vertical axis reflects the range of the data for libraries in all population groups--from quite small to quite large. In B and C charts, the scale range on the vertical axis is much smaller. For example, chart 13B shows boxplots for libraries in the less-than-1000 population (<1K) group to the 10,000 (10K) population group. In this chart, the vertical axis ranges from zero to 20,000.
Chart 13C displays boxplots for larger populations groups, from 25,000 (25K) to greater-than-500,000 (>500K). The vertical axis for this chart ranges from zero to 600,000, even though most libraries with service area populations greater than 500,000 surpass the top of this range. While boxplots for the four smaller population categories are fully visible in chart 13C, the >500K boxplot is hardly visible. However, the boxplot for the >500K population group is fully visible in chart 13A. (See Guide to Interpreting Boxplots for an explanation of the small diamond symbol.)
Another example of a chart with a close-up view is chart 14C, where about one-half of the >500K boxplot is visible. Again, to see the entire boxlplot symbol for the >500K group, the reader can refer to chart 14A where the chart's vertical axis accommodates the >500K subgroup completely.
At the top of each A chart appears a table of numbers, referred to here and in the NCLIS study as a sub-table. A sub-table lists descriptive statistics for each data element by library population group. Sub-tables are described further in the Guide to Interpreting Boxplots.