Table 1 (C)NMA models for (dis-)connected networks. \(n_{c} =\) number of subnetworks (connectivity components), \(n_{a} =\) number of intervention arms, \(k =\) number of studies, \(n=\) number of interventions, \(r=\) rank of the design matrx of additive CNMA model, \(\nearrow :\) increasing number of interactions in CNMA model, \(\searrow :\) decreasing number of interactions in CNMA model
Connected network | Disconnected network | |||
|---|---|---|---|---|
CNMA models | NMA | Separate NMAs for each subnetwork | ||
Additive model | Interaction models | |||
No. of interactions | none | 1, 2, 3, ... | all observed | all observed |
Model fit | often poor fit | \(\nearrow\) | often good fit | maximal fit |
Q | \(\searrow\) | minimal \(Q=\sum _{i = 1}^{n_c}Q_i\) | ||
Connectivity | good connectivity | \(\searrow\) | poor connectivity | minimal connectivity |
df | maximal \({df}=n_a - k - r\) | \(\searrow\) | \({df}=n_a - k - (n - 1)\) | minimal \({df}=n_a - k - (n - n_c)\) |