[R-meta] Confusion about how to use UN structure

[Simon Harmel]

Hello All,

I want to use the UN structure offered by metafor (rma.mv) and the
literature also supports correlating my moderator's categories in the
studies. But I get rho estimates of 1.00 which makes me think if there
is a problem with (a) my moderator itself, or (b) the way I fit the
model?

For example, a typical model I use is as follows:

rma.mv(eff_size ~ UN_MOD + x1 + x2 -1, V = V, struct = "UN",
                    random = ~ UN_MOD | study, data = data)

where UN_MOD has three categories (ex. type 1, type 2 and type 3).

I have had similar issues with the UN structure in the past, so any
thoughts are highly appreciated.

Simon