| Summary: | Optimizing Sample Surveys in Farming Systems Research. The collection of agricultural data, essential for Farming Systems Research, is expensive and the quality of the estimated statistics directly linked to the available budget. This paper shows how to calculate estimator precision of sample surveys from a small number of basic parameters, and how to compare it to a cost function for an optimal resource allocation. Basic parameters are calculated for key variables of the data base of the Farming System Research Team of Sikasso, Mali, and the estimator precision is derived for a number of options. It is shown that this precision can only be very modest with a minimal mean square error of 5 % to 10 %, for the typical small sample size (12 villages, each with 8 farms). Increasing the number of villages diminishes this error, but the effect of an increasing number of farms per village decrease quickly above six to eight farms per village. This type of surveys should therefore not be used for agricultural statistics. They do produce indicators of limited precision, but independend and readily accessible. They are better suited to analyze relationships between variables, to link causes wich constraints and reasons with technology adoption, and to follow major changes in the system's evolution.
|
|---|
