Hi

I am a GIS Technician at the USGS Coastal and Marine Geology. We used MGET "Fit GLM" and "Predict GLM" successfully this summer to combine ground truth observations, depth, class, and defined blocks to predict the distribution of benthic organisms. By successful, I mean that the coefficients in the "Fit GLM" exactly matched the coefficients we received when running the "R Formula" in "R Statistical package" independent of MGET.

Recently we reran our five taxa models and are getting "Fit GLM" coefficients that are not the same as the coefficient we got last time. This issue is occurring in both the exact same ArcGIS project (ArcGIS 9.3.1, Python 2.5, MGET .8a8, and R 2.6.2) where it successfully ran this summer and in a new MGET .8a21 ( I was unable to locate MGET .8a8. All other programs are the same: ArcGIS 9.3.1, Python 2.5, and  R2.6.2). As you can imagine, the predictions look different ranging from extremely different o slightly different?

Do you have any recommendations on what to try or why this could be happening?

Thank you for any advice,

Best regards

Nadine

Nadine Golden

U.S. Geological Survey

Pacific Coastal and Marine Science Center Santa Cruz, CA

(831) 427-4730