Demonstrate that two cohorts are comparable

[Claire Lewa]

Hi everyone, sorry for this basic question but I am a beginner.
I have two cohorts of patients, a bigger one (n=300) and a smaller one (n=100). I have the same data for both, except for the fact that the smaller cohort has one additional information (an additional test that the patients undergo during hospitalization). I need to show that these two cohorts are comparable. indeed, the aim of the analysis is to develop a model for predicting the development of a particular disease and assess the additional effect of the test, as compared to a regression model without the test and that i developed based on the larger cohort.
I need to show that the model prediction improves when the test is included, but not because of differences between the two cohorts.
I would be grateful if you could help.

[Maarten Buis]

The first step in your argument would be to discuss the design. If the smaller cohort was chosen by you and selected randomly, then you have a strong starting position. If the smaller cohort got the test as a response to how the disease was developing, then your starting position will be very weak. Still worrying but less so, woud be if the cohorts come from different sites.

After that you can look at the observed characteristics are similar in your treated and your control group. There are nice tools for looking at if groups are comparable on observed characteristics in tebalance, see help tebalance .

Then you need to compare your empirical findings with what you would expect based on your design. If there is a mismatch you need to come up with a convincing reason.

[Claire Lewa]

thank you!