1
$\begingroup$

Please, I have a question about the application of the logistic regression.

If we measure the glycemia (blood glucose) among the same subjects group in three time (measures were operated in 3 independent periods of intervention).

Then I would like to know if it is possible to apply the logistic regression to assess the association between glucose level and the different intervention points ?

my concern was to be sure that the three measures did not account as dependent variables (in another words I want to be sure that this situation did not constitute a violation of the the independence) ?

Thank you

$\endgroup$
3
  • $\begingroup$ Welcome to CrossValidated. I would ask you to give a bit more detail of the data you are handling. In particular, what is the question you want to ask to the data? Logistic Regression is a classifier, is not clear here (at least for me) what you want to classify. $\endgroup$
    – chuse
    Feb 4, 2019 at 10:33
  • $\begingroup$ Thank you for your reply. I would like to know if it is possible to use logistic regression to assess the association between high glucose (>2g/l) and the different study phase taking as reference group the first measure point. $\endgroup$
    – Dagdoug
    Feb 4, 2019 at 11:19
  • 2
    $\begingroup$ Although it's possible to use toothpicks to build a bridge over a river, it's of little practical use to know that. Similarly, although someone sufficiently clever might be able to think of a way to apply logistic regression to your problem, it's unlikely the solution will be of any use to you. Why not ask about the situation you actually confront, which is not about logistic regression but rather about how to analyze your data? $\endgroup$
    – whuber
    Feb 4, 2019 at 14:23

1 Answer 1

1
$\begingroup$

I have two concerns with this:

First, logistic regression is used when the dependent variable (here, glucose level) is categorical. Why would glucose level be categorical? Surely it varies over some range. Dichotomizing a continuous variable is almost never a good idea.

Second, if you measure the same people multiple times then you have, indeed, violated the assumption of independence. There are various ways to deal with this. I like multilevel models.

$\endgroup$
2
  • 1
    $\begingroup$ Peter: Thank you for your reply. 1#I used a recognized cut-off value for the glycemia but I can also analyse it as a continuous variable. 2# could you please help me to choose one of the multilevel models ( I am using stata) Regards, $\endgroup$
    – Dagdoug
    Feb 4, 2019 at 12:01
  • $\begingroup$ I don't know Stata at all, sorry. $\endgroup$
    – Peter Flom
    Feb 4, 2019 at 18:17

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.