Appendix C. Data and statistical tests on mean abundance and mean number of taxa in experimental patches.
Table C1. A priori contrasts of control vs. other treatments.
Response variable | ||||
Chthamalus stellatus | ||||
Patella aspera/caerulea | ||||
Encrusting coralline algae | ||||
Red filamentaous algae | ||||
Rivularia spp. | ||||
Taxa |
Notes: Data were analyzed using population-averaged generalized estimating equations (PA-GEEs), an extension of generalized linear models (GLMs) (Liang and Zeger 1986, Hardin and Hilbe 2003). These analyses compared the mean abundance of each response variable in control patches with that of disturbed patches over the entire duration of the study, by taking into account the temporal correlation between observations on the same experimental patch (mean values of the three replicate quadrats in each patch at each time of sampling were used as raw data in analyses). Tests used either a log-link (Taxa and Patella) or an identity (response variables other than Taxa and Patella) function and canonical distributions for the error terms. Temporal autocorrelation was modeled assuming a first-order autoregressive model.
E: estimated coefficient; SE: standard error.
All figures below illustrate temporal changes in abundance and number of taxa in control and experimental patches. Data are means (± 1 SE) calculated over three replicate quadrats pooled across three replicate patches in each experimental condition at each time of sampling.
Hardin, J. W., and J. M. Hilbe. 2003. Generalized estimating equations. Chapman and Hall/CRC, Boca Raton, Florida, USA.
Liang, K.-Y., and S. L. Zeger. 1986. Longitudinal data analysis using generalized linear models. Biometrika 73:13–22.