Malakar et al.

Statistical Analysis

A data matrix was constructed with habitat values and catch numbers by species for

each of the 40 sampling sites. Each habitat variable was represented by mean value

when repeated habitat measurements were made in different seasons Principal

component analyses (PCA) were performed on the 11 environmental variables

to reduce the dimensionality of the data and to examine initially the influence

of habitat characteristics. All variables are transformed log10(x + 1) before

data analysis to meet the assumptions of normality. Only principal components

with eigenvalues more than 1 were interpreted from PCA on the correlation

matrix of the environmental variables. Variable loadings more than 0.25 were

considered in structuring the stream using the method of Chatfield and Collins

(1980). A multiple analysis of variance (MANOVA) was conducted to compare

environmental variables. Relative abundance of each species was calculated from

the total number of individual sampled from all the sampling sites.

Canonical correspondence analysis (CCA) (ter Braak, 1986) was used to identify

the relationships of environmental variables with fish assemblage. Results of CCA

were tested with a Monte Carlo randomization method, which randomly reassigns

the values for the species data to the values for the environmental variables. Partial

CCAs were used to determine the variance explained by individual variables

after the removal of variables with inflation factors more than 10 (ter Braak and

Smilauer, 2002). All statistical analyses were conducted using the Multivariate

Statistical Package (MVSP) trial version 3.1 (Kovach, 1999).

Results and Discussion

Physicochemical Characterization

The PCA produced three axes that cumulatively explained 68.6% of the

environmental variation in sites (Table 2). The first axis had high loadings for

water flow, depth, temperature, turbidity and TDS. The second axis had high

loadings for temperature, dissolve oxygen, pH and TDS. The third axis had

high loadings for DO and conductivity. Eight variables (water flow, depth,

temperature, dissolved oxygen, pH, turbidity, conductivity and TDS) of eleven

total environmental variables had high loadings on at least one of the principal

component axis interpreted. In contrast, fine substrate, fine gravel, coarse gravel

did not have high loadings on any of the first three axis. TheANOVA on the habitat

variables indicated a significant difference in habitat structure among sampling

sites ( F =8.55, p <0.05).

138

International Journal of Bioinformatics and Biological Sciences : v.2 n.3&4. Dec. 2014