Figure 6

Relative bias and RMSE of the target concentration estimates in the presence of unequal partition size. When droplets or chips do not contain the same volume, bias is introduced. In the left panel, a boxplot shows relative estimates for 1000 simulated experiments at given concentration λ (copies/partition). The relative bias is calculated using a replicate based method as $\left(\widehat{\lambda }-\lambda \right)/\lambda$ for 1000 simulated experiments of 8 replicates. High concentrations show a downward bias. In the right panel, the associated root mean squared error $\mathit{\text{RMSE}}=\sqrt{\frac{1}{S}\sum _{s=1}^S{\left({\widehat{\lambda }}_s-\lambda \right)}^2}$ is shown, estimated as the square root of the sum of the relative variance and squared relative bias for S=1000 simulated experiments of 8 replicates. For a given concentration λ, this combines the errors as a result of the variance and the bias in a single number based on the results of 1000 simulated experiments. The best combination of accuracy and precision is achieved when the function hits its minimum.