0
$\begingroup$

I have run a large study looking at traumatic brain injury in patients I have conducted CT scans on patients very soon after the injury as well as neurocognitive testing and then repeated this at 1 year post-injury. At 1 year, I have also classified patients into "good" and "bad" outcomes based on their functional status. I also have control data.

I'd like to use a Fisher LDA to try and "predict" which patients will end up with bad outcomes from the initial CT scan / neurocognitive datasets.

What is the best way to go about this? That is, in an ideal world, would have two large data sets so could "train" on one and then run the prediction on the other

Is it OK to run the training on the 1 year outcome data and then apply this training to the data from the time of injury?

$\endgroup$
0
$\begingroup$

Assuming that you have data about each patients in both datasets I suggest taking only CT scans and neurocognitive test results from the time that occured soon after the injury, and the final status of the patient after one year. Then you can apply other techniques (such as CV or boot-strapping) to evalutate the classifiers performance.

$\endgroup$
2
  • 1
    $\begingroup$ Thanks. I have data from both time points. I am trying to use R to do this. Could you go into more detail for me? How should I run the classifier? $\endgroup$
    – montywaite
    Jun 24 '15 at 15:40
  • 1
    $\begingroup$ I recommed a caret package. Check this for a lot of useful examples. $\endgroup$
    – Khozzy
    Jun 24 '15 at 15:47

Your Answer

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

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