choice-metrics.com

Dear Ngene team,

I've done an efficient design and I obtained a d-error value of 0.00008. However, I've seen in couple of publications that their d-error is not usually that small.
Is it possible to have such a small d-error or there might be something wrong in my design?

Many thanks,
Eli

The D-error does not have a unit, so it is certainly possible that your D-error can be that small. It may for example depend on the units of your attributes. If you have used realistic priors, then the actual value of the D-error is not relevant, we can only say that the lower the better, but what is "low" is different in each model.

Thanks a lot for your response. It was really helpful.