Statistical Modeling for Soil Attributes

The study of nutrient patterns in soil is critical, and these investigations involve more multivariate existing factors. An empirical analysis was conducted to study the relationship between the soil characteristics, including the Potential of Hydrogen (pH), Electrical Conductivity (EC), Zinc (Zn), Sulfur (S), Phosphorus (P), Potassium (K), Organic Carbon (OC), Nitrogen (N), Manganese (Mn), Iron (Fe), Copper (Cu) and Boron (B) using principal component, Factor Analysis, and Canonical Correlation data reduction multivariate techniques. The objectives of the present studies are (i) to show how PCA, FA, and CCA are employed in reducing the dimensionality of the soil dataset without losing too much variability and (ii) to identify the correlation patterns that exist in the soil data, and (iii) to identify the soil data, which contribute most in the overall variance of the soil characteristics. Lastly, (iv) to show how nutrients in the soil can vary and to find the specific factors that make up the overall pattern of soil variation. The first PC was dominated by the soil characteristics N, Zn, pH, K, Fe, and Mn. The soil characteristics B, P, S, and Fe dominated the second PC. The third PC was dominated by B and P, while the fourth PC was dominated by the single feature Cu. The factor analysis of the first factor revealed significant negative loading on Zn and significant positive loading on Mn, pH, K, and B. The second factor has significant positive loadings for Fe and EC. The third component has significant positive loadings on Zn, S, P, and OC but has substantial negative loadings on N. A highly advantageous high Cu loading is seen in the fourth component. In the fourth component, Cu and OC are heavily weighted. The canonical redundancy for dependent and independent sets is 20% and 30%, respectively. According to the Stewart-Love canonical redundancy index, the initial linear combination of the X-set accounts for 39% of all the variance in the Y-set.