Our case studies using the Honduras databases have been able to screen dozens of socioeconomic ecoregional variables (from community to regional levels), and map clusters of communities for most efficiently targeting development assistance.

This output specifically addresses a problem that was brought to attention in the popular book, "How to Lie with Statistics". A more recent book, "How to Lie with Maps" by Mark Monmonier is as relevant and hopefully will become just as well known. The underlying issue is simply that manipulating the statistical sample can legitimize nearly any point of view. These tools offer protection against the unchallenged use of biased statistics, scenarios and maps.

From the literature it would seem that the scale issue is not well understood nor thoroughly researched. When this project was initially funded in 1997, there were few if any recognized texts or papers dealing with the issues of complexity and scale, and those that did, were not spatially explicit. Over the last four years, books such as Ecological Scale and Scale in Remote Sensing and GIS have appeared, but even in books such as these, the predominant theme seems to involve defining scale and talking about scale, rather than actually doing anything about scale.

The project aims to contribute to a series of techniques and tools that allow spatial data sets to be constructed and de-constructed in a generalized yet context sensitive manner. The outputs from such techniques can be explored and described through various user-defined levels, thus revealing spatial patterns and processes that are arguably more useful than raw data or standard representations. Potentially, complex hypotheses and models can be developed based on the improved understanding that such mapping techniques provide. Additionally, the opportunity to re-express the data at different levels - levels appropriate to different decision-makers - enables conflicts to be rapidly highlighted and the effects of a decision at one level to be visualized at other levels of organization. This is an essential step to address what Rhoades (1998) calls "scale wars" in project planning and execution, when each actor tries to get activities tied to the level with which he is comfortable with and have comparative advantage.

The identification of appropriate scales for analysis and prediction is a challenging problem. Although the factors producing scale dependent patterns may not be clearly understood, we have been able to create accurate and reliable descriptions of scale dependent patterns and processes to design data sampling procedures and test the accuracy and reliability of methods of prediction. There is clearly some way to go before scale effects can be fully understood and accommodated, but this research has aimed to be the 'next step' in that vital process.

The Spatial Data Exploration Toolbox includes tools to productively manipulate statistical samples in an exploratory "screening" of populations. Figure 3 illustrates an analysis resulting from the unique ability to delineate spatial "hotspots" at different scales using the Geographic Analysis Machine, one of the tools from University of Leeds that is part of the Data Exploration Toolbox. The figures are composite images of several analyses across scale, highlighting the problem areas in small communities (small spots) through to regions (larger patches).

Probably the greatest methodological challenge this Project has tried to address has been known for over fifty years, and amongst geographers is called the modifiable areal unit problem (MAUP). Symptoms of the problem are that the choice of both spatial scale and region of analysis will bias results and interpretation. There is no known cure. What is surprising is that so many books, papers and projects dedicated to issues of multi and across-scale analysis fail to address the issue, and in most cases, fail to acknowledge it's existence. This Project has gone beyond researching effects of MAUP, rather we have adapted available procedures to use the problem to advantage, by investigating alternative spatial representations of key variables, and by allowing census data to be re-expressed at new eco-regional levels (Accessibility Wizard).