indicator variable

[nurban]

Greetings,
We are working on a dataset with a lot of missing data. Different habitat types received different sampling which has left a lot of holes in our dataset. Specifically, we were wondering if there was a way for Presence to differentiate between different sites using an indicator variable. (ex. if 0=forest, 1=grassland then if 1 then the grassland covariate would apply and if 0=then not). If it does not recognize an indicator variable, we were curious as to how well Program Presence deals with missing data.

We appreciate any help on this matter.

[jhines]

If I'm understanding the model you want correctly, you could create a covariate which is zero for forest sites, and grassland-covariate for grassland sites. Then, build the model for psi as:

psi = beta1 + beta2*grass + beta3*newcovariate

where grass = 0 for forest, 1 for grassland,
newcovariate = 0 for forest, something for grassland sites

Then, psi would be constant for forest sites, and a linear function of the covariate for grassland sites.

Presence will warn you if you have presence-absence data with missing values for covariates, but will still try to run the model. When it comes time to compute a value based on a missing covariate, Presence will plug in -999 for the covariate (hopefully to make the results strange enough for you to investigate).

[darryl]

You might also want to (re)read the overview and tutorials in PRESENCE where most of this is covered....

Darryl