Hello,
I have done a few monte carlo matlab code alongs on youtube but none of them seem to apply to my situation. I am trying to use cancer biomarker levels in urine. I have a few other paraemters I use in a monoexponential growth model.

Monte Carlo picks value out of range. Determine detectable limit or what volume tumor you can detect with biomarker. Then, compare tumor volume to detectable level.

Calc stuff with monte carlo approach for each factor in the hori model (or what is published for HE4 + other biomarkers)

Generate random number in desired range for example;

Tumor size ( 0.4- 5mm)

Shedding rate (5 - 100) as example

Distribution with range? We can see if output follows normal/gaussian or uniform distribution. Once you have simulation, see if the results that match up or make sense. We can compare to real patient data to see if they agree.

I just feel over my head with this project. I don’t know how to set it up. Do you have ideas or resources I should check out?

Hi Emma. This is pretty far outside my area of knowledge, so I can only give some general pointers.

The way I like to think about Monte Carlo simulations is to imagine that you have real data from a real experiment. Thus, as much of the data should be as close as possible to the real thing. And then you have to pick the one (or maybe a few, but ideally one) data feature that gets randomly selected. I guess in your case that would be tumor size and shedding rate.

But first start by thinking about one sample, and having code to generate data for one sample. And then put that code inside a for loop to get lots of samples.

I realize that’s a fairly vague answer, but I hope it helps at least a little bit

I am trying to find code along for monte carlo and most of them on youtube are very specific for playing cards or “life guard” problem. I had bought two of your classes so already and was looking to see if you covered monte carlo in any other class? Maybe you know of another one?

I talk about MC sampling a bit in my stats+ML course. I understand that you’re looking for some boilerplate code that you can adapt to your application, but it sounds like your application is rather specific, which makes me think you might have to write it yourself.

Are there publications in your field that do similar analyses? If so, you could try emailing the authors to ask for their code. That’s often a good way to start.