Hi,
I am using DistLM in attempt to explain a proportion of the variation within my total bacterial community data set. I have a number of potential predictor variables from the soil parameters I measured.
I have conducted the DistLM and the marginal tests detected some significant predictors (i.e Total C, soil pH, and available Pi). These variables are significantly associated with the total bacterial community data individually, correct?
When I then look at the sequential tests, only Total C and available P have been included and included in the best model. If I re-run the model with the forced inclusion of soil pH it is automatically included in the sequential test and the best model but is this correct? Should I simply accept what the model suggests as the best model rather than manipulate it?
What does the sequential test try to achieve, the best combination of significant predictor variables from identified by the marginal tests?
Any advice is greatly welcomed
Kate