and effort. Once you are content with your conceptual model (figure 2.7), then you are ready to move on to the third phase of abstraction: the logical data model.
Figure 2.7 Sketching out the components of the conceptual model can greatly assist in identifying and organizing the information that will be necessary to complete a GIS analysis.
Defining terms or concepts is essential to developing a sound research project. This is especially true for names used to refer to phenomena. We advise you to conduct a complete literature review before beginning any type of research project to ensure you are aware of how different words are defined by others. It is likely that any concept you might want to name or investigate has a given name, and many concepts have multiple and, sometimes, competing definitions. The first step is to identify your concept, name it, and then define it. A literature review helps one to better understand the relationships that exist between different variables. Once you have identified your dependent variable, a literature review will further assist in explaining the relationships between the different variables. This is especially useful when you explore relationships between variables that have not been examined in the past. Once you have accomplished all of this, you are ready to plan your analytical approach.
Analytical approach: Phases of abstraction
When preparing an analysis in GIS, you must work through four distinct phases of abstraction, or modeling, in sequence:
1. Evaluate the real-world situation you intend to analyze.
2. Conceptualize in terms appropriate to a computer-based analytical approach.
3. Organize the logical approach to the analysis.
4. Implement the specific software steps.
Although many people conducting a GIS analysis tend to address the first three of these informally, taking the time to expand in some detail can greatly reduce missteps and dead ends as you carry out your analysis. In other words, taking a little extra time up front to plan the analysis process will yield benefits in the long run. We address each of these phases in the following sections.
Before you perform these four phases, keep in mind that, at this point, you are only considering the GIS aspects of the analysis. We assume that you have already determined the general nature of your study and that incorporating a spatial component through GIS will benefit your overall process.
Reality
Before initiating an analysis with GIS, it is essential that a GIS represents a model of reality—the computer does not understand reality. Although this may seem obvious, all too often, as people start to consider how they will work with their projects inside a GIS environment, they allow the GIS environment and their data to dictate their approach. Of course, knowledge of the data, the project, the disciplinary expertise, and local knowledge should drive the approach to the study.
Although it can be quite enticing, you should not allow the capabilities of a particular GIS software program or available data to dictate your analytic process. This is especially true when time, money, and personnel available to conduct an analysis are limited. Unfortunately, we cannot offer you a simple solution. Every project requires finding a balance between the real world that you are analyzing and the abstraction that is required to make that reality fit into a GIS. It is important to recognize that there is a distinction between reality as it truly is and reality as represented in the model. When assessing a GIS model or its result, keep yourself firmly planted in the true reality, where your results and decisions affect real people.
Beyond the issues of how the reality fits into the GIS, you will need to consider (as a precursor to this) what reality is for your study. Our discussion of reality isn’t philosophical but rather practical. Defining reality is not always quite as simple as it may seem. For example, if you are doing a survey of household income, where is the breakpoint between economic classes? How might this vary in different geographic regions? Are only dollars and cents relevant? What about bartered services? What about work performed in the home by a member of the household for no pay? As you can see, there are numerous possible components to how income might be assessed and analyzed.
As the analyst who is going to locate, collect, or use these data, how do you define your variables and classes? When looking for archival data, referring to the data dictionary and the metadata may help, if they exist. A data dictionary is a collection of information and details about the data that includes variable names. When collecting your own data, these are the kinds of issues you need to ponder and, in many cases, debate with colleagues, community members, or stakeholders. In short, defining reality is no small task, but it is by far one of the most important.
Logical data model
The logical data model adds specific processing steps to the conceptual model developed in the previous step, thus specifying the analytical procedures necessary to complete the analysis. That is, we want to add a workflow to the conceptual figure. This step can be represented graphically by creating a flowchart that specifies each step in specific detail. One unique characteristic about the logical model is that it does not include software-specific processing steps, only the logic behind them.
For example, in our prenatal clinic model, we know we need to extract specified socioeconomic data related to the prospective client population within our particular study area (city). The data source we have identified for this example is the US Census; however, because the census data contain much more information than we really need, we want to pare that down. We want to narrow the geographic extent and, via a database query, extract only those attributes we require. Graphically, this portion of the flowchart might look similar to figure 2.8.
Figure 2.8 A portion of a logical data model showing the steps necessary to reduce the full US Census dataset to the appropriate spatial and attribute components necessary for our study. The logical model first indicates a clip, which is a spatial operation to cut out only the required geographic area, much like a cookie cutter. The resulting dataset is then queried to obtain only those records or attributes needed in the research analysis.
In this example, we specify a logical sequence of steps but do not concern ourselves with how these steps are achieved in the particular software. It is important to have a sense of the possible processing options. Once we have developed a logical flow for the entire analysis, we reach our next evaluation point: determining the feasibility of accomplishing the necessary processing steps using available software tools (figure 2.9). Tools may vary based on the particular software licenses available for data collection, aggregation, and analysis as well as those that may be imposed by outside forces (e.g., field equipment used, versions of software).
Just as data may be brought into the GIS analysis from a variety of software environments, they can also be analyzed and processed in a variety of environments. You may determine that certain steps are better accomplished in a specific database program, a statistical software package, or a variety of other software options. In fact, it is not at all uncommon to incorporate multiple software programs and discipline-specific models into an analysis. For example, aspects of a project in ArcGIS may include data tables produced in Microsoft Excel or imported data from field equipment such as Global Positioning System data from Trimble Pathfinder Office. Many software packages commonly used in a GIS project are easily integrated with ArcGIS, either directly