Hi all. Some advice please!
We are interested in how sepsis influences the immune response in infants.
In our current study, we have taken repeated, weekly blood samples from infants on the neonatal unit who had developed sepsis.
At each time point, we measured several blood immune parameters.
In total we have approximately 20 infants enrolled, each with 10 blood samples.
Now looking at all 200 samples together I want to assess the influence of 'sepsis' (independent variable) on each individual immune parameter one by one (where the immune parameter is the dependent variable), controlling for confounders including sex, gestational age at birth and postnatal age in days (independent variables).
Does the below sound correct?
Using classical monocyte HLA-DR as an example of a non-normally distributed dependent variable:
In the fit model platform:
I select 'classical monocyte HLA-DR' as the 'Y' response variable
Then in 'random effects' I put the infants subject number- so that the model accounts for repeated sampling.
In 'fixed effects' I put the categorical variable Sepsis (yes or no) AND fixed confounder variables like birthweight and sex.
I also need to include a time factor, postnatal age, as the immune response changes over time. Do I include this in the 'fixed effects' section? Does this need to be as part of an interaction term? Is there a preferred model here e.g. GLMM?
Many thanks for any advice.
