Dear listers,
Could someone explain or point me in the direction to a routine for doing Cox´s calibration regression test; i.e. I have the predicted probabilities for death in -p- after having fitted a logistic regression. I´ve then generated the predicted log odds by -gen logop= log(p/(1-p))-. I was then going to fit a regression line (true logodds=alpha+beta*logop) but I was unsure as to what is meant by the true log odds and how I´d calculate this.
Also Is there a trick whereby I could generate a graph with observed probability on yaxis 1 and predicted probability on the x-axis, whilst simultaneously beneath yaxis=0 and above the xaxis showing a bihistogram containing the frequencies of dead and alive patients in opposite directions. I´ve been fiddling around with yaxis 2, but I´m only able to generate a graph with bars projecting along the positive y-dimension.
Regards,
//M
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
