0
$\begingroup$

If I wanted to analyze data taken of patients meeting pre-specified criteria from several hospitals through a linear regression (say how hypertension changes given age and sex), I think that a mixed/multilevel model would be appropiate, since data are non-independent and patients can be clustered by hospitals.

But, if I wanted to include ALL patients that meet pre-specified criteria from only ONE hospital, are my data still non-independent?

Many thanks.

Improve this question
$\endgroup$
3
  • $\begingroup$ What is it specifically that makes you want to focus on one hospital? $\endgroup$
    – Shawn Hemelstrand
    Commented Sep 13, 2023 at 11:45
  • 1
    $\begingroup$ We don't know your data, so we can't answer the question. A typical source of clustering in medical data is the surgeon/doctor treating the patients. $\endgroup$
    – Michael M
    Commented Sep 13, 2023 at 11:46
  • $\begingroup$ Thank you Shawn and Michael for your responses. I focus only on one hospital because we have patients enough to perform the study alone. We have only one doctor, so I think the only source of variability are the subjects themselves. $\endgroup$
    – user396498
    Commented Sep 14, 2023 at 6:25

1 Answer 1

Reset to default
1
$\begingroup$

Maybe. They aren't clustered by hospital, but they may be clustered by something else, such as doctor or ward or something like that.

However, when you focus on just one hospital, you lose a bit of external validity. The clustering by hospital hasn't gone away, you've just eliminated the ability to measure it, since you only have one. This might make it harder or riskier to generalize to other hospitals. If you want to generalize then you are assuming that your data are a random sample of all patients. This may not be correct.

Improve this answer
$\endgroup$
2
  • $\begingroup$ Hi Peter, many thanks for your response. Really, my data are not clustered in anyway (only one doctor and one hospital). This is the reason for which I was hesitant if I should use a normal linear regression or a mixed model grouping by subject. Probably the results are similar. $\endgroup$
    – user396498
    Commented Sep 14, 2023 at 6:20
  • 1
    $\begingroup$ You could only group by subject if you have multiple observations per subject, which would be a kind of clustering $\endgroup$
    – Peter Flom
    Commented Sep 14, 2023 at 12:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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