Table 2.

Accuracy results with the COVID-$19^{-}$ from the Costa Rican dataset, the higher, the better. LB stands for label balancing, with usual weight correction for the supervised model, and the proposed PBC for the MixMatch model. A total of $n_l=10$, $n_l=15$ and $n_l=20$ labelled observations were tested. Two data imbalance settings were tested, with $70\text{\%}/30\text{\%}$ and $80\text{\%}/20\text{\%}$. The sample mean $\overline{x}$ and the sample standard deviation $s$ are reported.

SSDL	COVID-$19^{-}$	COVID-$19^{+}$	LB	$n_l=10$		$n_l=15$		$n_l=20$
				$\overline{x}$	$s$	$\overline{x}$	$s$	$\overline{x}$	$s$
No	50%	50%	NA	0.871	0.039	0.912	0.049	0.951	0.025
	70%	30%	Yes	0.877	0.040	0.900	0.053	0.931	0.034
			No	0.877	0.040	0.924	0.056	0.931	0.044
	80%	20%	Yes	0.876	0.060	0.903	0.058	0.922	0.037
			No	0.876	0.079	0.907	0.072	0.938	0.035
Yes	50%	50%	NA	0.941	0.035	0.955	0.025	0.957	0.030
	70%	30%	Yes	0.955	0.027	0.947	0.035	0.950	0.029
			No	0.907	0.042	0.900	0.049	0.914	0.028
	80%	20%	Yes	0.957	0.025	0.964	0.021	0.960	0.020
			No	0.922	0.031	0.926	0.047	0.919	0.033