Examine and Select Environmental Variables

It is unlikely that all of the environmental variables will contribute usefully to the eventual model and there are often problems with multicollinearity.

The first stage is to simply run a canonical correspondence analysis off the ordination tab using all the environmental variables.

With the Dune Meadow data you will then read the following warning:

Click OK and ECOM will complete the analysis and open the output at the Multicollinearity tab.

Note the Variance Inflation Factor (VIF) values. A VIF above 10 indicates appreciable multicollinearity. In the above table it is clear that the Manure variable is highly correlated with a combination of other variables. Manure should probably be removed from the analysis. But first we can see if Manure is highly correlated with  any other single environmental variable.

Click on the Scatter Plot tab. Select Manure in Plot (x Axis) and each of the other environmental variables in turn in the y axis menu box. If you activate the box you can use the up and down arrows to swiftly move through the variables. Make sure you have selected Env. in the Plot radio button or you will only be offered the species!

The scatter plot of Manure against SF indicates a Pearson correlation between these variables of 0.72. Evidently the standard farming variable and Manure are related. This increases our suspicion that Manure or possibly SF should be removed from the final analysis.

We will not remove any variables from the present analysis but for future reference this is done using the Deselect Row in the Working Environmental data tab.

It is often useful to identify the top 1, 2 or 3 environmental variables for inclusion in the final analysis. This can be done using the Variable Selection drop-down menu.

The next stage is to interpret the results.