Figure 5From: Sweet tweets! Evaluating a new approach for probability-based sampling of TwitterEstimated regression slopes from predicting \(F_{RD}^{T}\) from \(\hat{F}_{RDMS}^{T}\) by region, method and size. The regression models each use the frequency estimates from the 8 topics of interest across the 38 days of the experiment and are fitted separately for each region, method and size combination and fix the intercept at 0Back to article page