By: David N. Barton (NINA), Youssouf Cisse (IER), Bocary Kaya (Millenium Villages Project), Ibrahima N.Diaye (IER), Harouna Yossi (IER), Abdoulaye Diarra (IER), Souleymane Keita (IER), Amadou Dembele (IER), Daouda Maiga (IER), Graciela M. Rusch (NINA), Anders L. Madsen (HUGIN)
The paper can be accessed here.
8 March 2016
The article discusses the potential of BNs to complement the analytical toolkit of agricultural extension. Statistical modelling of the adoption of agricultural practices has tended to use categorical (logit/probit) regression models focusing on a single technology or practice, explained by a number of household and farm characteristics. Here, a Bayesian network (BN) is used to model household level data on adoption of agrosilvopastoral practices in Tiby, Mali. We discuss the advantages of BNs in modelling more complex data structures, including (i) multiple practices implemented jointly on farms (ii) correlation between probabilities of implementation of those practices, and (iii) correlation between household and farm characteristics. The paper demonstrates the use of BNs for deductive reasoning regarding adoption of practices, answering questions regarding the probability of implementation of combinations of practices, conditional on household characteristics. As such, BNs is a complementary modelling approach to logistic regression analysis, which facilitates exploring causal structures in the data before deciding on a reduced form regression model. More uniquely, BNs can be used inductively to answer questions regarding the likelihood of certain household characteristics conditional on certain practices being adopted.
To use the model deductively click on a value for at least one of the Farm(er) characteristics and observe the change in the adoption probabilities across the Silvopastoral practices. To use the model inductively, click on at a value for at least one of the Silvopastoral practices and observe the change in the likehoods of different Farm(er) characteristics having adopted particular combinations of practices. (The model monitors are best viewed with the Browser in FULL SCREEN mode)
David N. Barton, Youssouf Cisse, Bocary Kaya, Ibrahima N.Diaye, Harouna Yossi, Abdoulaye Diarra, Souleymane Keita, Amadou Dembele, Daouda Maiga, Graciela M. Rusch, Anders L. Madsen (2016) Diagnosing agrosilvopastoral practices using Bayesian networks. Agroforestry Systems, pp 1-10. DOI 10.1007/s10457-016-9931-1.
Useful references for those interested in BBN include:
Kjærulff, U. B. and Madsen, A. L. (2013) Bayesian Networks and Influence Diagrams: A Guide to Construction and Analysis. Springer, Second Edition.
For further details on the paper: David N. Barton (NINA)
For further details on the use of Bayesian networks and web deployment of models contact: Anders L Madsen (alm(at)hugin(dot)com)