06-08-2018 02:16 AM - edited 06-08-2018 07:31 PM
This is Kelly and I need someone's help>"<
Currently, I am working on developing a risk equation for diabetes and stroke.
After I got my equation through cox-regression(proc phreg) and also there were roc curve and c -statistics.
However, I would like to compare my equation to other equations which have been published to see whether my equation is more precise.
That is, I would like to ask if there any procedure to use my cohort to test the existed model to calculate c-statisitcs?
Attachment is the study method I found may be similar to mine. The author mentioned using SAS, however, not sure what procedure did they use................
Kindly give advise to me...........Thanks a lot!
06-08-2018 05:48 PM
you are estimating survival time ie length of life? i'm not too familiar with this, but i am aware of an old paper with frank harrel's name on it: https://onlinelibrary.wiley.com/doi/pdf/10.1111/j.1532-5415.2000.tb03126.x they compared several models using discrimination and calibration. But I'm more familiar with risk modelling with a dichotomous outcome... then there are good methods and sas macros
06-09-2018 02:40 AM
Thanks for your kindly reply!
However, not sure how to apply the risk equation:
risk_score = 0.0634*age + 0.0897 *m_b_hbA1C + 0.5314*log10( m_U_ACR)
5_year_probability = 1-0.9707**(Risk_Score-4.5674)
into my dataset to look into the c-statistics and ROC curve for this equation?
Looking forward your kindly reply