Which method is best for testing model Fitness while using Hierarchical logistic regression using large data sets?
It would be helpful if you described your regression methods in more detail. Best, David Booth
Hello Girum,
It depends on what you're meaning by fit. If you'd like to know how sample specific a model might be, you could (with sufficient numbers of cases) split your sample into two (or more) subgroups, derive a model using one of those subgroups, then apply that model to the other subgroup(s) to see whether it performs comparably.
If you mean via measures like AIC or BIC, or AUC, then these can be determined for logistic models (whether simultaneous or hierarchical/stagewise).
If you mean internal consistency, such as Hosmer-Lemeshow applied to subsets of the model-building group, this is available in many logistic regression software packages.
If you mean something else, then perhaps you could elaborate your query.
Good luck with your work.
Thank you very much guys for your help. And sorry for not my case clear. To make it clear i am trying to assess the impact of biomass fuel use (explanatory variable) on birth size using a large data set. I am using hirarceal logistics regression. In the first stage I tried to test all the explanatory variable with child factors and in the next stage I tried to to assess the explanatory variable by including the child and maternal factors. In the final stages i I have included the child, maternal and socio-demographic factors to assess the impact of biomass fuel use on birth size. What I wanted is in each stage which goodness of fit test is best to use because different litratures suggest that using Hosmer-Lemshow test for large data sets is not recommended. if using Hosmer-lemshow test is not a good idea what are my options?
And do I have to worry about the goodness of fit test in all the stages?
Once again thank you for your help.
- Badges
- Science topic