Because our data structure involves measures obtained
from transactions that are divided into different suppliers,
the data are collected in a p-variate vector Yts (where p is the
number of variables considered in the model and subscripts t
and s denote transactions and suppliers, respectively).
Through MSEM, we decompose the individual data Yts into
a between-supplier (or supplier-level) component YB ¼ ¯Ys,
and a within-supplier (or transaction-level) component
YW ¼ Yts ¯Ys. Since the two components are additive and
orthogonal (Searle et al., 1992), the observed total measurement
vector YT decomposes as YT = YB + YW.