Questions to be addressed

Results from the different studies are reported by Romney et al. (2005).

The data were difficult to analyse. For example, the longitudinal study was set up as a participatory study because it was considered important to understand the benefit of the concentrate feeding strategy when placed in the farmers' hands.

Following good biometric practice the researchers organised treatment and control groups to allow formal statistical inference. But the farmers did not do as expected.

Volunteer farmers did not follow the guidelines, and some of the control farmers, when seeing what the volunteer farmers were doing, did likewise. Thus the results of the study were driven by the actions of the farmers and the strategy for analysing the data had to be adjusted accordingly.

In view of the complexity of the data collected by Romney et al. (2005) we have decided to investigate the adequacy of two of the approaches to the statistical analysis and to examine whether a more rigorous treatment of the data would lead to the same or a different interpretation of the data.