I DO need help for this classification problem. The data is the classical Fisher's Iris one.
And the multinomial logit model was used for finding a classification rule.


The problem of this model is that for setosa/virginica, the numerical difficulties occured and the coefficients of logit(setosa/virginica)
actually failed to converge.

The output I enclosed here includes the Logistic Regression Table and Descriptive Statistics for four measurements by Species.
U may somehow notice that the poor fit for logit(setosa/virginica) may due to the distinct difference between these two species.

My task is to, basd one the poor fitted model and these descriptive statistics, find a classification rule for the iris variety.

Hope someone here could help me. Many many many thanks sin....

[此贴子已经被作者于2008-10-8 21:18:55编辑过]