Cross Validated
2026-11-16 15:30 UTC
By user13317
AI-113-20261116-social-media-4f89d96d
Conditional Mean of a Mixed-effect Model
Admittedly dumb question, but if I have a mean model with the following form, $$ \mu = \text{logit}^{-1} \left( \beta_0 + b_i \right), $$ The intercept represents the average value over individuals, $i$ , on the link scale. If I calculate the following, $$\text{metric} = \text{logit}^{-1} \left( \hat{\beta_0} \right), $$ is it correct to say that the metric represents the conditional mean of a median individual ? Setting $b_i = 0$ and assuming $b_i \sim \text{Normal} \left(0, \sigma_i^2\right)$ . If I wanted the conditional mean of an average individual (not sure this is possible), what would I need to do? I think the distribution of the individual random effects would be logit-normal on the response scale. Just trying to better understand the metric, and how that would differ from a marginal mean.
Admittedly dumb question, but if I have a mean model with the following form, $$ \mu = \text{logit}^{-1} \left( \beta_0 + b_i \right), $$ The intercept represents the average value over individuals, $i$ , on the link scale. If I calculate the following, $$\text{metric} = \text{logit}^{-1} \left( \hat{\beta_0} \right), $$ is it correct to say that the metric represents the conditional mean of a median individual ? Setting $b_i = 0$ and assuming $b_i \sim \text{Normal} \left(0, \sigma_i^2\right)$ . If I wanted the conditional mean of an average individual (not sure this is possible), what would I need to do? I think the distribution of the individual random effects would be logit-normal on the response scale. Just trying to better understand the metric, and how that would differ from a marginal mean.
Full article content could not be extracted automatically. Read the original below.
Source:
Cross Validated
· stats.stackexchange.com