This dataset contains the time-dependent effective reproduction number Rt of the COVID-19 epidemic in Finland for the period Feb 27th 2020 - August 14th 2020. The reproduction number was estimated using the daily new coronavirus cases in the infectious diseases registry of Finland with an unscented RTS smoother and the SEIR epidemic model. The SEIR model was defined as \(S(t+1) = S(t) - \frac{R_t(t)}{T_i} I(t) \frac{S(t)}{N} + \omega_{S}\\ E(t+1) = E(t) + \frac{R_t(t)}{T_i} I(t) \frac{S(t)}{N} - \frac{E(t)}{T_e} + \omega_{E}\\ I(t+1) = I(t) + \frac{E(t)}{T_e}-\frac{I(t)}{T_i} + \omega_{I} \\ R_t(t+1) = R_t(t) + \omega_{R_t}\) with S, E, and I being the susceptible, exposed, and infective populations, Rt the time-dependent effective reproduction number, and ωs are normally distributed random numbers with zero mean and small (1e-8) variance except for ωRt whose variance was set to 0.00025. The exposed state duration was set to Te=3 days and infective state duration Ti=5 days. The measurement model (from the SEIR state to observed daily new cases) was defined as \(P_\text{new cases}(k) \propto \mathcal{N}\left[ \alpha \frac{R_t(t-5 days)}{T_i}I(t-5\text{days}) \frac{S(t-5\text{days})}{N}, \sigma_\text{new cases}^2(t) \right]\), i.e., via a time-delayed gaussian process whose variance was computed by the observed daily new cases with a 7-day rolling mean. Here the detection rate α is set to 10%, but has no significant affect on the computed R-values. The data is a CSV-format file with the columns date = ISO 8601 date string Rt = maximum a posteriori estimate of Rt Rt_lower50 = lower limit of the 50% credible interval for Rt Rt_upper50 = upper limit of the 50% credible interval for Rt Rt_lower90 = lower limit of the 90% credible interval for Rt Rt_upper90 = upper limit of the 90% credible interval for Rt