0
$\begingroup$

If we are doing a cohort study and investigating whether a variable has an effect on an outcome (say tumor size on nausea), and we run a regression (or chi2) with nausea as a dependent variable and tumor size as an independent variable, would we strengthen our study by including patient characteristics in a multiple regression?

So include their gender, age, weight etc, effectively "controlling" for these variables.

To me it seems it would but knowing statistics, there's always a catch.

$\endgroup$
1
$\begingroup$

If you are interested in these variables and their effects then you should include them. If they are not relevant to your research question then don't include them.

These variables could "strengthen" your model by improving model fit, but again whether you should do this depends on your goal.

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ Well if my research question is as simple as "Does tumor size affect the rate of seizures in brain cancer patients", would including patient characteristics strengthen the results? I don't really know if age, weight or ethnicity has anything to do with seizure rates, I can check with a chi2 or a single variable regression, maybe they do. If they don't in the single variable regression, should I just exclude them? If they do, should I include them? Again, I'm not sure if they pertain to the research question, maybe they do. $\endgroup$ – Paze Nov 26 '19 at 10:10
  • $\begingroup$ @Paze The question is whether this is of interest to you, not whether these variables have an effect. If testing whether patient characteristics have an effect on your outcome is not your primary goal, but maybe is still of interest, then you can easily include them and get some additional information out of the model. However, if they are absolutely not of interest, because this information is useless to you, then you can ignore them, because even if they improve model fit, this is irrelevant for you. $\endgroup$ – user2974951 Nov 26 '19 at 10:39

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.