The stochastic distribution editor (see the picture below) is used to parametrize stochastic distributions, e.g. duration parameters for activities, interarrivaltimes for events and stochastic based process variables. You can open the editor via the property editor of the corresponding element. See also a complete list and description of available distributions.
Figure 1: Stochastic Distribution Editor 
Every time you select and parameterize a distribution a curve will be drawn, which predicts the value distribution in the assigned space.
Note that:
Advanced Options
Advanced Options 

No negative number generation 
Check this if you do not want the simulation to generate any negative numbers from the selected distribution. This value is checked by default (true). It cannot be changed, if the distribution is used to generate values for a time consumption. But you do have the possibility to generate negative numbers where it makes sense (see this help's topic on Process Variables for more information). 
Antithetic number generation 
This option allows you to generate the exact opposite values of the actually drawn values. The definition of an exact opposite (antithetic) value depends on the distribution type. Every distribution implements an own algorithm to determine this value. E. g. you select the Discrete Uniform Distribution with the parameter setting: MinValue= 2, MaxValue=6 for the duration of an activity. If you check this option and if we assume the first (actually) drawn value is 2, then the first result will be 6, because it is the antithetic value to 2 in this particular distribution. This option can be useful, if you want to analyze a specific activity. Just like setting a manual seed for a certain distribution, this is a way to obtain significantly different generated numbers for exactly one BPMN element that uses any kind of distribution. It is unchecked (false) by default. 
Manual seed instead of autoseed 
By checking this option, you can set the seed for exactly this distribution to a value, that you determine. This way the distribution will not be dependent on the global seed. This approach is useful when you want to analyze one particular activity (or anything else using a distribution) in several simulation runs. The point is to leave all the other options constant and changing only the seed for this particular distribution. 