+ Reply to Thread
Results 1 to 4 of 4

Thread: Validation for prognostic marker

  1. #1
    Points: 63, Level: 1
    Level completed: 26%, Points required for next Level: 37

    Posts
    4
    Thanks
    0
    Thanked 0 Times in 0 Posts

    Validation for prognostic marker




    Dear All,

    I am completely lost in the area of statistics. I have a dataset of 265 cancer patients with a novel protein marker as well as the known clinical markers. I read somewhere that I can split my dataset to two and use one as derivation and the other as validation cohort.
    However, I have no idea how to model based on derivation cohort and see if it will fit the validation cohort.

    The marker scores are continuous so I tried to find a cutoff point which I can dichotomize my dataset. I chose the median. It shows a pretty significant KM analysis in derivation cohort but not in validation one. Also, cox regression multivariate analysis with other clinical factors does not give me anything significant.

    Can you guys help me how to validate what I find in one cohort to the other?

    Thanks a lot

  2. #2
    Omega Contributor
    Points: 38,289, Level: 100
    Level completed: 0%, Points required for next Level: 0
    hlsmith's Avatar
    Location
    Not Ames, IA
    Posts
    6,992
    Thanks
    397
    Thanked 1,185 Times in 1,146 Posts

    Re: Validation for prognostic marker

    Is your sample size bigger enough given the models, to split into two. Depends on your purpose, but typically you randomly split datasets. Design a model on one then test it on the other. Is this not what you are trying to do?
    Stop cowardice, ban guns!

  3. #3
    Points: 63, Level: 1
    Level completed: 26%, Points required for next Level: 37

    Posts
    4
    Thanks
    0
    Thanked 0 Times in 0 Posts

    Re: Validation for prognostic marker

    Thanks for your reply.

    Yes that is exactly what I wanna do.
    How big is enough? My whole dataset consists of 260 patients. 45 deaths. Do you think it is enough to be split?
    Also, what do you mean by modeling? So far, what I did, I split the dataset into two and used cutoff finder to find a cutoff point for my marker and applied that point to the validation cohort. The problem is that it is dependent on many other clinical factors. How can I correct for those?

  4. #4
    Points: 63, Level: 1
    Level completed: 26%, Points required for next Level: 37

    Posts
    4
    Thanks
    0
    Thanked 0 Times in 0 Posts

    Re: Validation for prognostic marker


    Hello again,

    OK! I am reading a lot of stats and regression these days. Here is where I am facing a problem:

    In a book called multivariable analysis, I read that I can split my dataset into two (which I did), generate a model in one (Which I did with cox regression and found all the coefficients and their SE ) and use it to estimate survival in my validation set and compare the estimated and observed probabilities of the outcome. My question is how I can perform such an analysis, ie estimate the probabilities of the outcome in a new dataset using the same cox models? And then perhaps it would be good to plot the estimated and observed ones on a graph and get the R squared to see if the model fits well.

    I have no idea how to do all of those but I am very excited about what I (think!) am getting into!

    Please let me know what you guys think. It is probably very easy for you guys.

+ Reply to Thread

           




Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts






Advertise on Talk Stats