 |
 |
I like to calculate model R-squares because it is an intuitive measure of how well a model explains the data and, more importantly, because it is something that the consumers of my analyses, natural resource managers, understand. For a spatially-explicit, hierarchical model I've been working with in BUGS I calculate the deviance R-squared as: 1-(Model Deviance/(n-k-1)/(Null Deviance/(n-2)), where n is the sample size and k is the number of model parameters. Heretofore I was using pD, the number of estimable parameters, in place of k, having derived pD from the DIC menu in WinBUGS1.4. In some of the models pD is very large, substantially reducing the model R-squared. In one model, the "naive" Deviance R-squared is 1-(Model Deviance/Null Deviance) or 1-(8078/20811) = 0.61. Adjusting for sample size considerations, n = 1840 and pD = 1077, the adjusted Deviance R-squared becomes 0.06, an order smaller. The number of covariates in this model is 7, so the inflation of pD is due strictly to the spatial structuring of the model. In another similar model with a different response, pD is only 40, and thus has considerably less influence on the calculation. I haven't seen people use pD in place of k and so I don't really know if this is a valid practice, so your comments would be appreciated if you have any. I'm wondering if I should stick strictly with the number of model covariates rather than include all of the hyperparameters. If you can recommend another formulation of R-squared that might be more appropriate, I'd be interested in that too. Thanks much, Wayne E. Thogmartin, PhD Statistician (Biology) - Avian/Habitat Modeling USGS Upper Midwest Environmental Sciences Center 575 Lester Avenue Onalaska, WI 54650 608.783.7550 Ext 42 [log in to unmask] ------------------------------------------------------------------- This list is for discussion of modelling issues and the BUGS software. For help with crashes and error messages, first mail [log in to unmask] To mail the BUGS list, mail to [log in to unmask] Before mailing, please check the archive at www.jiscmail.ac.uk/lists/bugs.html Please do not mail attachments to the list. To leave the BUGS list, send LEAVE BUGS to [log in to unmask] If this fails, mail [log in to unmask], NOT the whole list