Using Stata 11 & higher for Logistic Regression
Model fit: How often is the model right?
Create a classification table
% predicted correctly
page 6: “To get the equivalent of SPSS’s classification table, you can use the estat clas command (lstat also works). This command shows you how many cases were classified correctly and incorrectly, using a cutoff point of 50% for the predicted probability.”
Stata FAQ: How can I perform the likelihood ratio, Wald, and Lagrange multiplier (score) test in Stata?
Model significance: Use likelihood ratio test to compare two models (m1 nested in m2):
logit hiwrite female read estimates store m1 logit hiwrite female read math science estimates store m2 lrtest m1 m2
If the test is significant, the bigger model is a better fit.
Correction: This webpage is about linear regression, not logistic regression. But still, the measures described are useful – just choose wisely. There are two things to do for logistic regression:
Assumptions for logistic regression: linearity, independence of errors, multicollinearity.
Identify outliers & influential observations.
From this webpage:
The following table summarizes the general rules of thumb we use for these measures to identify observations worthy of further investigation (where k is the number of predictors and n is the number of observations).
|Cook’s D||> 4/n|
VIF: > 10
tolerance: < 0.1
Definition and interaction:
Visualizing Main Effects and Interactions for Binary Logit Models in Stata:
Logistic Regression with Stata (Xiao Chen, Phil Ender, Michael Mitchell & Christine Wells):
Applied Logistic Regression (David Hosmer and Stanley Lemeshow):
Stata Programs for Data Analysis: