|Classification and Regression Trees, Cart TM - A user manual for identifying indicators of vulnerability to famine and chronic food insecurity + disk (IFPRI, 1999, 56 p.)|
This manual has laid out the fundamental theory underlying Classification and Regression Tree (CART) analytical techniques, and also explained how such techniques can be applied in practice. Concrete examples were presented from research at IFPRI. This research has explored the potential of CART to provide a less subjective framework for the selection of famine risk indicators and determine the relative importance of such indicators in explaining vulnerability across years and regions in Ethiopia.
The theoretical exposition and the results from applied CART analysis suggest that this methodology offers considerable potential for assisting in the analysis of large and complex data sets. CART also offers a transparent, "objective" methodology upon which planners can base their decisions.
That said, CART should be seen as one tool that can be used, in conjunction with others, for analyzing data, assessing risk, and planning development. The technique is extremely data-intensive and, hence, labor-intensive (in terms of the time an analyst spends collating, preparing, and analyzing the data). What is more, there remains a need for further research into the definition of appropriate benchmark indicators (such as the "population in need" figures used here), against which multiple variables can be tested. In the short run, the choice of indicators will most likely be driven by data availability. But in the longer run, such choices should be made as a result of assessments of the reliability and sensitivity of alternatives.
Further exploration of the gains and drawbacks inherent in CART are therefore encouraged, and not just in relation to research on food security. As IFPRI and others have demonstrated, CART can be usefully applied to a wide range of uses.