IJA_2026v16n5

International Journal of Aquaculture, 2026, Vol.16, No.5, 293-314 http://www.aquapublisher.com/index.php/ija 303 Figure 7 Seasonal variation in major ions and nutrient concentrations (chloride, sulphate, NN) and Phosphate in the upper catchment of the Sankh River 3.5 Statistical analysis of seasonal variability 3.5.1 Seasonal variation in physico-chemical parameters (MANOVA) PCA is necessary before applying Multivariate Analysis of Variance (MANOVA) when classifying data where the number of parameters equals or outweighs the observations. In this case, even having only nine samples from three seasons and a set of water quality variables is complex enough that a standard MANOVA would produce a rank-deficient variance-covariance matrix. This deficiency would make the test statistically invalid and would forestall the computation of test statistics. By implementing PCA first, this collapses the original correlated parameters into a small number of orthogonal Principal Components (PCs). These factors detect maximal data variance while minimizing the number of variables to a level feasible within the limitations imposed by available degrees of freedom. However, there are two major justifications for choosing MANOVA over multiple ANOVAs: the inflation of Type I error and keeping correlations between variables. Performing separate ANOVAs increases the possibility of detecting significant differences based on random chance alone, an effect known as multiplicity. More critically, water quality parameters like pH, temperature, and dissolved oxygen are inherently interdependent. MIC calculates the correlation between chemicals, and MANOVA does so across all seasons simultaneously as one big multivariate vector. It allows for the detection of subtle shifts across all parameters at once, information that individual ANOVAs (which would analyze each variable more or less independently) were likely to miss.

RkJQdWJsaXNoZXIy MjQ4ODYzNA==