3
$\begingroup$

In this experiment cows were fed with two different types of grass (Treatment) and we measured the weight gain over time, keeping track of the individual's ID. Sex of the individual was also taken into account. I used GLMM to analyze the data, using (1+Sampling | ID) to account for the repeated measure design.

Weight gain ~ Treatment * sex + Sampling + (1+Sampling | ID)

My question is, how do you describe in a paper the random effect that accounts for the repeated measures? So far I have: "treatment, sex and sampling were included as fixed effects, as well as the interaction term between treatment and sex. Sampling and the individual's ID were included as random effects". But I think that last part is not accurate.

Thanks

$\endgroup$
3
  • $\begingroup$ Can you say more about sampling? I assume these are the time points when you measured weight. How many time points? Do you treat sampling as discrete or continuous? Were all cows measured at the same time points after starting the experiment? Also, what exactly is the outcome variable? The weight or difference in weight between consecutive measurements? Did you measure weight before the experiment started? $\endgroup$
    – dipetkov
    Sep 7, 2023 at 18:30
  • $\begingroup$ @dipetkov Correct, sampling is the time points when cows were weighted, which happened every 2 weeks for 6 months. The variable is treated as continuous. All cows were weighted at the beginning of the experiment and they were all weighted at the same time during every sampling event. The response variable is the cumulative gain in weight since the beginning of the experiment. $\endgroup$
    – M.Pinilla
    Sep 7, 2023 at 18:49
  • $\begingroup$ It seems to me the random slope assumption is strong, esp. if the cows were adult cows. You have 12 measurements per cow and assume the relationship with time is linear. Why did you include the random slope? Have you made plots of the average/individual growth to evaluate this choice? $\endgroup$
    – dipetkov
    Sep 7, 2023 at 19:04

1 Answer 1

2
$\begingroup$

With your model formula:

Weight gain ~ Treatment * sex + Sampling + (1+Sampling | ID)

...you have described the fixed effects fine, but the random effects description should be expanded a little:

Something like: "Treatment, sex and Sampling were included as fixed effects, as well as the interaction term between Treatment and sex. Random intercepts were fitted for ID and random slopes were fitted for Sampling"

I would also suggest looking into the the comments by @dipetkov about fitting random slopes for Sampling (that should be a new question).

$\endgroup$

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.