TABLE 4.
Model parsimony test.
| Model | EP | Δ Fit | Δ df | p | AIC | Δ AIC | Compare with model | Fit units |
|---|---|---|---|---|---|---|---|---|
| Cholesky | 20 | 65,863 | 0 | −2lnL | ||||
| Reverse causation | 19 | 136.15 | 1 | < 0.001 | 65,997 | 134.15 | Cholesky | −2lnL |
| Forward causation | 19 | 4.597 | 1 | 0.033 | 65,865 | 2.597 | Cholesky | −2lnL |
| No cause | 18 | 136.152 | 2 | < 0.001 | 65,995 | 132.152 | Cholesky | −2lnL |
| Feedback loop | 20 | 135.3 | 0 | < 0.001 | 65,998 | 135.3 | Cholesky | −2lnL |
Note: All models were tested against the Cholesky model, which was also the best fitting model (Figure 1). AIC, Akaike information criteria; df, degrees of freedom; EP, estimated parameters; lnL, log‐likelihood.