*2.3. Data Analysis*

Data normality and homogeneity have been tested using the Shapiro–Wilk test [41] and the Levene test [42], respectively. Non-normal data have been transformed into Log(x) and Log(x + 1), and the remaining non-normally distributed data have been analyzed by non-parametric tests.

The density variability in adult colonies and recruits over the years 1998 to 2013 has been analyzed via the non-parametric Friedman test [43]. Moreover, the population size/age structures over this period have been compared by multidimensional scaling (MDS): in the resulting configuration of points (*n* = 9, years of sampling), the closeness of one year's sample data to another's provides a measure of similarity. The MDS analysis has been performed using the free software Past4.03 and the Euclidean index of similarity has been used.

Differences in percent cover of TSH and occupied substrate between the two sites (*n* = 12 at Tinetto Island and *n* = 11 in Tinetto Shoal) sampled in 2013 have been tested via the *t*-test [44]. Percent cover and density differences in adult *P. clavata* colonies between sites have also been analyzed via the *t*-test, while those of the recruits with the Mann–Whitney test.

The relations between adult and recruit densities and between these data and free substrate have been tested by R<sup>2</sup> best fit coefficient, linear Pearson's coefficient and the *t*-test. The relations between recruit densities in the same quadrate on 2 subsequent years have been explored on the scatter plot: the density of one year was plotted against that of the following year in the same fixed plot (2007 vs. 2008 and 2009 vs. 2010), finally a linear trend has been calculated.
