I have a sample data of 583 type 2 diabetes patients and want to calculate the 5 year incidence probability of an event for every patient. Variables which were collected are time to an event variable censored at 5 years, event status and 38 independent parameters. Cox regression was giving non significant results for all the variables so elastic net regression specified for a cox family was applied. The following commands in R were used:

x<- cbind(Age,Sex,Tobacco,Alcohol,Energy,HbA1c...)

After obtaining the beta coefficients, the risk score (RS) is determined from RS = B1*X1 + B2*X2 + ... + Bn*Xn. Next, I want to calculate the 5 year probability of the event. Molnar et al. (2017) used the formula "1 - S(5)EXP[RS]" and Yang et al. (2007) used the formula "1 - S(5)EXP[RS - mean of RS]".

My questions are:

  1. How to calculate the survival function S(t) after applying glmnet? Is survfit command an option? Yang et al. states that S(t) is the survival function over t years when the risk score takes the value of its mean. How can this be determined?

  2. Which formula is the better choice to use? I can't find any background on the formula developed by Yang et al.

Any help will be highly appreciated.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.