Meta-analysis, missing estimates of standard error in the outcome

We are running a meta-analysis using brms, with effect size (hedges’ g) as an outcome (mean and se). However, for some studies we were not able to reconstruct the uncertainty of the effect (se).
Is it possible to specify that the se is missing? something like

y = c(0.1, 0.2, 0.3)
y_se = c(0.1, 0.05, NA)
f = data.frame(y,y_se)
brm(y|mi(sd(y_se))~1, f)

which gives me: “Stan does not support NA (in noise) in data”

It is not possible at this point. What would you think is a good approach at handling the missing se?

1 Like

Would the model here be pertinent?

I think that if you had the sample size but not the SE or SD, you could use partial-pooling to inform on the missing SD and thereby permit studies with that info missing to still contribute to the meta-analysis, no?