From: RFisher on 17 May 2010 15:39 Hi everyone, I'm working with a dataset on nest use of a small grassland songbird. What I've done is to develop models of nest use (0=nest, 1=random) using 5 explanatory variables (i.e., variables related to vegetation structure) in a logistic regression framework. Using AIC I've found my top models and have used model-averaging to determine model-averaged parameter estimates and associated standard errors. I've then used these model-averaged parameter estimates to determine predicted probabilities of nest use based on each variable individually, while holding the other variables constant (at average values). For example, I wanted to calculate the predicted probability of nest use at vegetation heights of 5cm, 10cm, etc.. But I can't figure out how to calculate SEs around those predicted probabilities. Is there a specific formula, is there a way to do this directly in SAS? Note, I've had to calculate the predicted probabilities by hand since I can't seem to figure out how to do it in SAS using model-averaged parameter estimates and not simply the parameter estimates from the "best" model. I hope that is sufficient information! Thanks, Ryan
From: Bminer on 20 May 2010 17:58 On May 17, 3:39 pm, RFisher <ryanfis...(a)sasktel.net> wrote: > Hi everyone, > I'm working with a dataset on nest use of a small grassland songbird. > What I've done is to develop models of nest use (0=nest, 1=random) > using 5 explanatory variables (i.e., variables related to vegetation > structure) in a logistic regression framework. Using AIC I've found my > top models and have used model-averaging to determine model-averaged > parameter estimates and associated standard errors. I've then used > these model-averaged parameter estimates to determine predicted > probabilities of nest use based on each variable individually, while > holding the other variables constant (at average values). For example, > I wanted to calculate the predicted probability of nest use at > vegetation heights of 5cm, 10cm, etc.. But I can't figure out how to > calculate SEs around those predicted probabilities. Is there a > specific formula, is there a way to do this directly in SAS? Note, > I've had to calculate the predicted probabilities by hand since I > can't seem to figure out how to do it in SAS using model-averaged > parameter estimates and not simply the parameter estimates from the > "best" model. > I hope that is sufficient information! > Thanks, > Ryan Ryan, No help but very interested. How did you calculate the parameter coefficients and SE under model averaging? Then do you create a CI using a normal approximation? Love if you can share how you calculated them. Brian
|
Pages: 1 Prev: Data input glitch Next: SAS created Excel files unreadable under Win 7 |