Home | Login | Logout | View Survey/Sites | Add Survey/Sites | My Account | UserID:

ANOVA Application to the Reef Check Dataset

Overview

ANOVA applied to the Reef Check survey database provides a means of assessing whether certain types of factors are influencing the numbers of key indicator organisms that are being monitored on reefs, and whether significant differences are being observed over time and in space. This in turn may suggest clues as to the detailed ecological mechanisms involved while also providing general indications regarding the health and overall status of reefs. The Reef Check WRAS ANOVA system supports a wide range of options and provides a generic tool for formal analysis of the Reef Check survey database. But whatever the list of available factors for selection and filter parameters, ultimately the user should choose combinations of categorical variables and data subsetting criteria that are meaningful. Users also should interpret results in a careful and sensible manner. All of the available functionality and these considerations are reviewed below.

Categorical Factors

The principal factors of interest to Reef Check are spatial and temporal in nature. People using the system should thus be able to perform analyses to assess whether significant differences in measurement variables (ie. abundance counts for each of the 43 fish and invertebrate indicator species) are observed between places and/or over time. Given the global nature of the Reef Check dataset, analysis based on multiple spatial scales is both highly desirable and possible. Spatial factors or scales over which analyses can be conducted range from Regions, to sub-regions, to Countries or combinations of these factors. With respect to time, mutliple time scales could also have been considered. However, comparisons between years seemed most appropriate given the frequency of data collection and the length of the available time series. Therefore, temporally users can only look for potential differences between years. Depending on precisely what the user specifies in the (spatio-temporal) Factor and Subfactor fields of the ANOVA web-form and as the measurement variable (indicator species - one can select only one of these for a given analysis), one could for example compare the abundances of groupers observed between reefs and years and assess whether significant differences occur. By repeating the analysis for the same categorical factors, but altering the choice of observed variable (say to Parrotfish), one could examine whether results are sensitive to the choice of particular types of indicator species. Or one could decide that time may not be important, and look not only at whether indicator species abundances vary significantly BETWEEN countries, but also consider different reef systems WITHIN countries, and so on and so forth.

Filters for Restricting Data Used in ANOVA

The Reecheck ANOVA system supports the use of filters to select particular subsets of data for use in analyses. These have been included because it is not necessarily the case that all users will want to conduct analyses on the full dataset, which is global in scope and spans several years. A researcher in the Caribbean, for example, interested only in making comparisons of reef indicators between countries and sites of his region, must have a means of excluding non-Caribbean reefs from the analysis. Furthermore, it is not necessarily the case that a result observed for a particular area will be observed when other areas are included in the analysis or whether different/additional spatial scales are considered. Similarly the same researcher may only be interested in making comparisons between particular years (eg. before hurricane and after) and not for the full available seven-year series. A time filter therefor exists in addition to filters allowing the specification of Regions, Sub-regions, and Countries for inclusion in particular analyses.

Methodological Caveats & Data Transformation

As mentioned previously, ANOVA is a parametric method with relatively stringent data requirements. Unfortunately, however, abundances of organisms (which is what population census data such as those of Reef Check are about), invariably violate both aforementioned assumptions of normality and heteroscedacity. For this reason, the Reef Check WRAS ANOVA system analyses are based on log-transformations of the raw abundance data (more precisely: log(Value+1), +1 being applied because log(Zero) is undefined and zero abundances do occur and are important to include in analyses). Such a transformation is standard, and makes valid the application of ANOVA to the type of data that are available here.

Considerations Regarding Interpretations

Much care must be employed when interpreting ANOVA results and making ecological inferences for population census data (as opposed to controlled experimental design type data). The fact that an analysis shows that significant variations in grouper abundance occur between years does not necessarily imply, for example, that reef habitat and health has deteriorated; fish population sizes are naturally extremely variable in both space and time. As such, significant differences spatially and/or temporally do not necessarily say anything about the health of the population or reef unless one can show in what direction changes are occurring and possible linkages to likely explanatory variables. If ANOVA detects that significant differences do occur between years AND this is part of a systematic trend of decreasing abundances correlated with changes in habitat availability then you may have shown something interesting impact-wise. Otherwise the ANOVA may simply be resolving natural variability. As with any statistical analysis for ecological data, some expertise and understanding of the systems and data involved is required, and over reliance on any one type of analysis ill advised. The combination of XY plotting and ANOVA in the WRAS system is particularly important in this regard.