A4 Logistic Regression Models
- Formulate and solve a model to run logistic regressions
- Use i=1,...,N observations (e.g. N=100 or 1000)
- Use m=1,...,M randomized features x[i,m] (e.g., M=5)
- Use realized dependent variables y[i] (0,1 binary)
- Compute the optimal beta-coefficients (e.g., using the minos solver)
- Compute the AIC0 of the "Null-model" with one feature only (i.e., a constant x_1=1)
- Compute the AIC of the model with all M features
- Compute the McFadden Pseudo R^2 := 1 - AIC/AIC0
Hint: Extend the linear regression model (linear_regression.txt)
An objective is not needed.
Solve the M-dimensional nonlinear system of equations (see regression lecture, slide 27)