Hey there Reddit,
I'm working om my master thesis which as of yet is taking way too much time. My study centers around a model which uses a relatively unknown method to predict mortality of ICU patients. The model has already been created by someone else (or actually, a bunch of different people over different studies) and I'm going to take that model, a. test if the results of the previous researcher hold up in a different dataset, b. change some aspects of the model in terms of how it deals with the dataset (splitting of obersvations in training / testing and dealing with the point-in-time aspect of the observations) and c. combine a yet-to-be-collected dataset with 4 other datasets from different hospitals around the world in order to create a more robust model.
I'm having GREAT difficulty however translating the work which I want to / am going to do to proper research questions. As a whole the writing of my lit study / proposal / thesis is taking ages and I i have a mentor who is quite busy.
What are research questions in validation studies? or is just "what is the external validity of the model" enough? also how do i differentiate between my research questions when it comes to external validity of the model as-is (Trained/tested with a single dataset) and what I want to do (train/test a model with multiple datasets) or isn't there a difference?
Hope you can help!
tl;dr: Going to validate and improve an existing predictive model and cannot seem to find proper research questions.
[link][comment]