marcinkosinski / coxphsgd Goto Github PK

View Code? Open in Web Editor NEW

7.0 7.0 1.0 139.36 MB

Stochastic Gradient Descent log-likelihood estimation in Cox proportional hazards model

R 100.00%

coxphsgd's People

Contributors

Stargazers

Watchers

Forkers

luyajun01

coxphsgd's Issues

an example with coxphSGD and rkafka streming

Summation over Risk Set

When you compute the cumulative sum over the observations in the risk set, do you use all observations in the risk set or a randomly selected subset of observations? It appears that you are using all observations for this summation, but I may be missing something.

times variable

during CRAN CHECK

* checking R code for possible problems ... NOTE
coxphSGD_batch: no visible binding for global variable ‘times’
prepareBatch: no visible binding for global variable ‘times’
Undefined global functions or variables:
  times

animation to blog post with gradient descent

http://vis.supstat.com/2013/03/gradient-descent-algorithm-with-r/

Does coxphSGD do random sampling?

Hi. May I ask how coxphSGD take "part/batch of samples" ?
In my case, I have two groups. First one contains >100 patients, second contains ~10 patients. Is coxphSGC feasible for such comparison?