Error Message when I Evaluate models with Bootstrap

[Chris Bowman-Prideaux]

Hi Bruce!

I am trying to evaluate my selected model with bootstrap analysis. The randomization setting is off. My model has 5 predictor variables. When I try to run the analysis, I get the following error message:

"Too many possible models for resampling; reduce number of predictors."

Does the statement about reducing the number of predictors refer to the number of predictors in the model or the predictor matrix?

I tried looking up the information in the Help menu. My system has a menu table of contents and index, but when I try selecting a topic, no information appears in the display window. Am I missing a help file somehow or is this a problem with Hyperniche v2.30?

[Chris Bowman-Prideaux]

I removed one of the predictor variables from my model, making it a model with 4 variables and the bootstrap analysis is running. Does anyone have a suggestion on criteria I should use to eliminate a variable from a model? Also, how does that affect the interpretation of the original 5-variable model when I am reflecting on it in my results and discussion?

Thanks,
Chris

--
You received this message because you are subscribed to the Google Groups "HyperNiche and NPMR" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hyperniche+...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--

Cheers,

Chris Bowman-Prideaux

[Bruce McCune]

If the Help document is not displaying, it is possible that Windows is blocking access (because it is a downloaded file that is an executable). First, check in the folder containing the HyperNiche software for the file HyperNiche2.chm. If it is not there, then there was a problem with the installation and you should redo the update, being sure to install to the same folder, overwriting the old files. If the file is there, you can make sure that Windows will let you use it by right clicking the file name, then Properties | General tab. At the bottom right if you see an "Unblock" button, click on that. If there is no Unblock button then that probably isn't the problem. If that isn't there, one other thing to try is the way you launch HyperNiche. Close HyperNiche, go to the icon from which you launch it, then right click, go to Compatibility tab, and check "Run this program as an administrator."

About the "Too many models..." message. That means that the number of models you are asking it to evaluate is more than 10^9 or you are asking it to do a randomization test with more than 4 predictors. The latter is actually an arbitrary cutoff having to do with speed of execution. The computation time goes up exponentially with the number of predictors and in our experience, more than 4 gets you into the glacial zone with a reasonable number of randomizations.

Bruce McCune

--

[Bruce McCune]

I would suggest removing the predictor with the lowest sensitivity. Alternatively you could remove the predictor that entered the model last. In most cases that would be the same predictor. Unfortunately, passing a randomization test with a 4-predictor model is not the same as for the 5 predictor model. I suppose I would re-examine how important that fifth predictor is. In most cases when you have that many predictors, additional predictors give you only minute increases in the cross-validated fit.

Bruce McCune