Data transformation
Prior to multivariate and spatial analyses for the forest soil
around lead–zinc mine, the normality of all heavy metals was
checked (Chen et al., 2008). The parameters like skewness, kurtosis
and the significance level of Shapiro–Wilk’s test for normality (S–K
p) are presented in Table 3. It was observed that none of the heavy
metals passed the Shapiroe–Wilk’s normality test (S–K p > 0.05)
before data transformation. Their variables were strongly skewed,
with the value higher than 0. Their kurtoses were also very sharp
governed by the fact that the majority of samples were clustered at
relatively low values. The logarithmic transformation and Box–Cox
transformation were used to normalize the original data.
All heavy metals except Pb (S–K p=0.49 > 0.05) still could not
pass the normality test even after the logarithmic transformation.
In comparison to logarithmic transformation, all the Box–Cox
transformed data passed the normality test except for Cr and Cu.
These metals had such a high kurtosis that the logarithmic and
Box–Cox transformation methods could not create absolute
normal distribution. The data suggested that the concentrations
of these two elements were not high but had strong randomness
suggesting an influence of the natural sources and distribution. As
the Box–Cox transformed data for Cr and Cu did not pass the
normality test, very low skewness implied that its transformed
distribution was very close to normalization. The Box–Cox
transformed distribution for Cr and Cu was closer to normal
distribution as compared to raw data distribution and lognormal
transformed (Fig. 2)