Table 6 Ablation analysis results across models for originally correct causal arrows. Note that the 1st experts’ proposal has the most negative values of all, meaning that removing those arrows individually produces less likely models.
From: Enhancing social science research on cyberbullying through human machine collaboration
Model | Removed arrow | \(\Delta\)LL |
|---|---|---|
1st experts’ | Age \(\rightarrow\) Hours_Internet | \(-\)14.19 |
Gender \(\rightarrow\) CB | \(-\)32.21 | |
Age \(\rightarrow\) CB | \(-\)48.73 | |
Hours_Internet \(\rightarrow\) CB | \(-\)51.94 | |
Sexual_Orientation \(\rightarrow\) CB | \(-\)26.31 | |
BS | Gender \(\rightarrow\) Family_Communication | \(-\)19.05 |
GTT | Gender \(\rightarrow\) Family_Communication | \(-\)19.05 |
NO TEARS | Age \(\rightarrow\) CB | \(-\)12.08 |
Hours_Internet \(\rightarrow\) CB | \(-\)15.53 | |
Sexual_Orientation \(\rightarrow\) CB | \(-\)10.86 |
