Post on 06-Oct-2015
adjust variables to dichotomous variablesgap can range
ols
coefficients and std errors
B = (xx) (xy)
sigma = sigma sqrd (xx)^-1
sigmaB = sqrt(diag(s)
GLSB = (xU^-1X)^-1 * xU^-1y)(xI^-1x)^-1 * xU^-1y(xx)^-1 *xys
identity matrix I multiply matrix by 1 is the same thing(xIx)^-1 *xIy
U^-1
We weight by their variance. Weigh high variance lower, low variance high.
S = sigma^2(xU^-1x) ^-1
ProbabilityE(y) = p = y/nV(y) = p(1-p)