Need help in estimating random intercept value for each group for reporting into a journal summary table???

[Rahul Raoniar]

I would like to estimate a random intercept linear regression model based on several covariates, and also like to investigate the influence of a group variable (site: from where the data were gathered). The site variable consists of 8 locations.

After model fitting with mixed command in Stata 17 (with site as group variable), it generated two tables. One for fixed effects and another for random effects.

As per literature the basic equation used for random intercept model is given by y_ij = beta₀ + beta X_ij + u_j + epsilon_ij, Where, beta₀ is the constant and beta₀ + u_j represents the intercept for group j.

After literature review, I found that many researchers reported the beta₀ + u_j for each group, (here my group variable is site).

So, I thought that stata might have any command that directly report it, but I'm unable to locate it even after reading the stata 17 documents for multilevel models.

My question is: How one can compute the intercept value for each group? I would like to report the site wise intercept than a global intercept provided by the fixed effects table.

Any help regarding this will be appreciated.

[Clyde Schechter]

predict u, reffects
gen intercept = u + _b[_cons]

That said, 8 sites is a pretty small sample of site-space and the estimates you get for these site-level intercepts will be pretty noisy. I wonder if it wouldn't make more sense to do this as a one-level model with i.site as a covariate instead.

[Rahul Raoniar]

Initially, I thought of adding the site as a covariate but based on past publication experience (where reviewers asked to fit a mixed effect model), this time I have estimated a random intercept regression model.

Yes, you're right that even a one-level model could be appropriate in place of a mixed effect model, as it explained only 12.7% of variance that was not accounted by the fixed effect model.

Thank you for the solution.