Common components and specific weights analysis (CCSWA) is a relatively recent multiblock statistical method that constitutes an extension of principal components analysis (PCA) in the case where different sets of quantitative variables have been measured on the same set of individuals. We described in this thesis the principle of CCSWA and its application in R software on real data to analyze farmers' perception of land degradation and soil erosion in northern Benin (West Africa). The data considered bear on 5 sociocultural groups and variables are linked to the causes of land degradation (dataset 1), soil erosion factors (dataset 2), land use practices against soil erosion (dataset 3) and techniques of improvement of the soil fertility and crops productivity (dataset 4). On these datasets, we also applied PCA in order to show the improvement of CCSWA compared to PCA. The results of CCSWA showed that the common component q1, opposing Djerma to Haussa farmers according to local perception of land degradation and soil erosion, expressed 60.4 %, 45.3 %, 10 % and 73.5 % of the total inertia of datasets 1, 2, 3 and 4 respectively. Djerma farmers think that land degradation is due to erosion, agricultural settlement and wildfire. Run-off and slope are the main soil erosion factors according to them. They also think that crops productivity can be enhanced by using plows and carts. Regarding Haussa farmers, deforestation is the main cause of land degradation, whereas the soil type is the main soil erosion factor. Against soil erosion, they set up stony lines and use manure and household rubbishes to improve the soil fertility and crops productivity. The common component q2 explained 5.4 %, 30.8 %, 70 % and 9.4 % of the total inertia contained in datasets 1, 2, 3 and 4 respectively and opposed Dendi to Djerma farmers about local perception. Dendi farmers acknowledge animal stamping and soil type as the main soil erosion factors and practice fallow to improve the soil fertility and crops productivity. As regards Djerma farmers, they cover their lands and till orthogonally to the normal ow of water in order to overcome soil erosion. Globally, the results of CCSWA and PCA are almost the same but the improvement that CCSWA brings is the knowledge of how different datasets cooperate to form the common components.
Date of publication:
RUFORUM Theses and Dissertations
Agris Subject Categories:
Romain L. Glele Kakai