using random numbers to model medical statistics?

Post Reply New Topic

Okay, a medical study typically has an experimental group and a control group. And one ethical way to do it is for the control group to get standard good care and the experimental group to get standard good care + new care. But people may not always be able to receive double medicine. So, there might be some choices to make. And hopefully, patients and participants will be kept fully informed and freely make choices whether to participate or not.

And once medical researchers run the study, they typically use various statistical packages to say how likely the observed difference between the experimental and control groups is due to chance alone. There may be an additional way.

We can look at how much variance is in the control group, how much spread. And then we can say, okay, what if we ran this 100,000 times? Say we randomly take a subset of twenty persons from the control group, and we see how their average compares to the average of the control group of the whole.

1) Has anyone heard of anything like this?

2) And not necessarily instead of currently used statistical methods, but as something used in addition to current methods, does this sound like a good idea?

I've not done much work with stats, but I've heard a few terms such as Monte Carlo simulation, Random Walk or a Brute Force method. Both terms make it sound quite crude, but I think if you google them you should be able to find they are a useful way of looking at probably outcomes. Certainly in engineering simulation in early concept work this can be a useful method to create a model to "test" a solution to see if there is merit in developing the idea.

It's always worth a look. It sounds like you know what you want to do, you just don't know the term for it so if you search on these I think you might find the support you are looking for.

Thank you for the leads.

Years ago I read, not in a science or mathematical book, but in a book on philosophy and ethics that it was important to compare within-group variance to between-group variance. For example (my example), if someone says 'women are better at telephone customer service than men,' that may or may not be true in some average sense. But there are plenty of men who are steady eddie, personable, and friendly at phone customer service, as well as plenty of women who are just not very good at it at all. So, in hiring decisions, we come to the not very remarkable conclusion of evaluating people as individuals. And on social justice terms, we might also come to the conclusion that a healthy economy would have more good jobs than it currently has.