How to handle biases

[andreped]

When performing the statistical analysis it is important to set it up in such a way that the p-values mean what we want them to mean.

To avoid wrong interpretation, we have to adjust for different biases and use the right model and hypothesis test(s).

One bias that comes to mind is whether we should filter away or include the patients that had surgery. There are quite few of them, so ti is probably a good idea to remove them. However, by removing them we introduce a selection bias, meaning that there is a population of tumours which we are not including in the statistical analysis, which could have been relevant.

It all depends on what the hypothesis questions are. It would be ideal to know this before choosing which modelling technique to use.