The Sampled Results distribution provides a flexible way to create a non-parametric distribution using a list of sampled (observed) results. GoldSim generates a CDF by sorting the observations and assuming that a cumulative probability of 1/(Number of Observations) exists between each data point. If there are multiple data points at the same value, a discrete probability equal to N/(Number of Observations) is applied at the value, where N is equal to the number of identical observations.
In addition to providing the list of observations, you also specify whether or not GoldSim should extrapolate at the ends of the data set. If the Extrapolation option is cleared, a discrete probability of 1/(2*Number of observations) is assigned to the minimum and maximum values. When the Extrapolation option is selected, GoldSim extends the generated CDF to cumulative probability levels of 0 and 1 by extrapolating existing observations.
A simple example (using 200 data points) is shown below:
Typically, a Sampled Results distribution would only be used with a fairly large number of data points (50 or more). With fewer data points, in most cases, it would be more appropriate to fit the data to a known parametric distribution.
Note: You can copy and paste the
entries for a Sampled Results distribution from a spreadsheet. To do so, select
the column of data points in the spreadsheet, and copy them to the clipboard.
Then place the cursor in the first field in the GoldSim dialog, and paste (by
pressing Ctrl+V).