Research strategy

Before contemplating a study to evaluate a 'before/after' intervention the researcher needs to be satisfied that a sufficient period of time can be allowed for the study.

The length of period that will be needed will depend on the likely impact of the intervention and to what extent formal statistical analysis is required.

An immediate, clear, major impact will not require formal statistical analysis, just a careful evaluation of the likely levels of influence of alternative seasonal or management factors. A few years, however, are needed to confirm the impact of an intervention such as tsetse control, especially on animal performance variables.

To summarise, when considering a research strategy for assessing the impact of a treatment the researcher needs to:

• decide whether it is possible to plan a study with parallel sets of intervention and control farmers,

• if not, decide what lengths of baseline control and intervention periods will be necessary and the likely effects of uncontrollable factors over time,

• decide on an appropriate time interval definition for the observational unit,

• consider other data that may be useful to collect - such a rainfall or other meteorological or environmental data - that may correlate with variations in the response variables under investigation.