Relative ecological indicators are frequently used tools in vegetation analyses. Despite their ordinal
nature, it has been shown that average indicator values can characterize an area well, and can provide
useful ecological information. Several different averaging methods have been tested againstthe indicated
environmental parameters, but only very slight differences could be found between their reliability. Different
statistical tests, including parametric and non-parametric tests, are also often applied on relative
ecological indicators. Similarly to the weighting methods, there are several ways to provide source data
for the tests from raw indicator values but the possible differences in the reliability of the resulting statistical
layouts have never been looked at. In the present study we have chosen the Hungarian adaptation
of Ellenberg’s indicator for soil moisture as a model system and examined a total of 8 different statistical
layouts. Raw indicator values were obtained from vegetation surveys of 16 appropriately chosen sites and
were processed in two fundamentally different ways. In the first approach, average indicator values were
calculated for each sampling quadrat of the sites and these averages were used as source data for ANOVA
tests. The calculation of the averages was carried out in four different ways according to the weighting
methods. In the second approach, site specific species lists were compiled using the quadrats of each site
and the raw indicator value populations deriving from these lists were analyzed with Kruskal–Wallis
tests. Again, four weighting methods were used, but instead of averaging, the indicator value of each
species within a site was repeated as many times as its weight required. Finally, the reliability of each
method was assessed by comparing the results with the actual soil moisture relations of the sites, determined
with physical measurements. According to our results, it can be said that false positive results
are rare with any type of the methods but the amount of false negative results varied among the methods
considerably. The most reliable method was the Kruskal–Wallis test when performed on frequency
weighted raw indicator value populations. This method could best reproduce the original soil moisture
relations and could yield the most convincing p-values; therefore we can recommend using this method
in studies where sets of relative ecological indicator values are intended to be compared with statistical
tests.
©