An exploratory data analysis showed that the values of microcystin concentrations in D. polymorpha tissues (including zero values) had considerably right-skewed distributions with several very apparent outliers. Therefore it was decided to apply log-transformation for the data, in order to minimize the effect of outliers. Since general assumptions of the parametric analysis methods (Shapiro–Wilk normality test, p < 0.014; Fligner–Killeen test of homogeneity of variances, Vorinostat ic50 p < 0.05) were not met neither before nor after transformation applied, the further data analysis was performed using non-parametric methods. To compare the results of microcystin concentrations

gained by two different analysis methods (ELISA and PPIA), the non-parametric Wilcoxon signed rank test was applied. Non-parametric Kruskal–Wallis test was applied to compare microcystin

concentrations found in muddy and sandy sediments in 2008. The multivariate effects of the studied factors – time of sampling (combining year and month), sampling site and mussel size – on the concentration of microcystins in the mussel tissues were analyzed by statistical program PRIMER 6 & PERMANOVA (Anderson, 2001 and Anderson, 2005). PLX4032 The test-statistic is a multivariate analogue to Fisher’s F-ratio and is calculated directly from any symmetric distance or dissimilarity matrix. p-Values are then obtained using permutations. In the current study the Euclidean distance similarity

measure was used to construct the similarity matrices. The statistical differences between the factor levels were assessed by four-way PERMANOVA with “time of sampling” (6 levels), “size” (3 levels) and “location” (2 levels) as factors. The permutation of raw data was used as this method is recommended in the case of relatively small sample size ( Anderson and Robinson, 2001). When a factor and/or interaction was identified as significant (p < 0.05), post hoc PERMANOVA pair-wise tests were conducted to detect which levels were responsible for significant differences. Multiple regression was applied to the log-transformed microcystin concentrations data from the zebra mussel tissues with mussels size and sampling time as explanatory variables. Microcystin concentration in mussel tissues varied from values below the detection limit Arachidonate 15-lipoxygenase to 139 ng/gDW when measured with ELISA test and from values below the detection limit to 284 ng/gDW when measured with PPIA. Although the pair-wise comparison of the two applied sample analysis methods, ELISA and PPIA, has shown no significant differences in the obtained results (W = 1.13, p = 0.26), in concentrations higher than 10 ng/g DW PPIA tended to give greater values. In order not to lose any data and minimize the undesirable bias, the results of the both tests were considered in the multivariate analysis as response variables.