1
$\begingroup$

I divided my dataset into Test and Validation (50-50 split).

I ran glm function (link=binomial) on Test dataset and got the parameter estimates.

How do I score the Validation dataset based on these parameter estimates (beta) that I got from Test dataset. I know it has something to do with apply () but I am not sure. please advise.

Improve this question
$\endgroup$

1 Answer 1

Reset to default
1
$\begingroup$

Split sample validation is highly unstable if $n < 20000$ and is not recommended. If you do have an enormous test sample you can use the R rms package val.prob to do the validation and its help file will give you some pointers. But better to look at the bootstrap and the rms validate and calibrate functions.

Improve this answer
$\endgroup$
2
  • $\begingroup$ Thanks Frank, can you give me reference to the article that states " validation is highly unstable if n<20000 "....Thanks Frank. $\endgroup$
    – bison2178
    Commented Feb 17, 2014 at 7:37
  • 1
    $\begingroup$ You can run a simple simulation or refer to my book Regression Modeling Strategies. $\endgroup$
    – Frank Harrell
    Commented Feb 17, 2014 at 13:26

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.