All numbers in the results are rounded to 2 digits after the decimal point.

For modelling the trend of the attacks is used a linear model, $$ y = a t + b + \epsilon$$ $$ \epsilon \sim N(0, \mu)$$ where y is the number of attacks, t is the time step, a is scale, and b is offset. The time-step is determined by the resolution, which can be hour, day or week.

The parameters of the model are fit according to $$ a = \frac{\sum_i (y - \mu_y)(t_i - \mu_t)}{\sum_i(t_i - \mu_t)^2} $$ $$ b = \mu_y - a \mu_t $$ $$ \mu_t = \frac{1}{n} \sum_{i=1}^n t_i $$ $$ \mu_y = \frac{1}{n} \sum_{i=1}^n y_i $$ Then the uncertainty of the scale is estimated by assuming Gaussian distribution of the errors, according to $$ \sigma_a = \sqrt{\frac{\sum_i (y_i - a x_i - b)^2}{(n-2)\sum_i (x_i - \mu_x)^2}} $$

For further details, please check linear regression.