Table 3.

Correlation between F1-binary, Negative Agreement, F1-macro, and F1-micro with class imbalance. This is done for two different networks and the “Ones’ control group. As can be seen in this table, different network architectures have similar correlations. For example, F1-binary has a high correlation across both networks, and F1-micro has a negative correlation across both networks. This suggests that the specific network does not have a high impact on the metrics obtained and the class imbalance plays a larger role.

Metric	Correlation
	Network 1	Network 2	Ones
F1-binary	0.9758	0.9489	0.9912
NA	−0.8359	−0.9031	0.9874
F1-macro	0.7570	0.6349	0.9961
F1-micro	−0.2656	−0.3118	0.9913