Calculation Method

By default, Crystal Ball tries to fit a normal distribution to the forecast values. You can enter a significance level to specify the threshold below which the assumption of normality is rejected. The default level of 0.05 translates into a 95% confidence that a rejection of normality will be correct. Other significance levels typically used are 0.01, 0.025, and 0.1, which translate into 99%, 97.5%, and 90% confidences, respectively.

If normality is rejected, Crystal Ball will then either calculate the metrics directly from the forecast values (the default) or, if you choose, perform a best fit to select the most appropriate continuous probability distribution from which to calculate the metrics.

The normality test and non-normal best fit (if normality is rejected) use the goodness-of-fit test and distribution selection that is set in the Forecast Window tab of the Forecast Preferences dialog (opened by choosing Preferences, then Forecast in the forecast window).

Before you choose to calculate from the best fitting distribution if the distribution is not normal, consider that:

You are not guaranteed of attaining a good fit to the forecast values, and

The fitting process may take a long time depending on how many simulation trials you are running.

Note:

In unusual circumstances, it is possible for the normality test to fail and the best fitting distribution still be a normal distribution, or for the normality test to pass and the best fitting distribution be non-normal.

Alternatively, you can select the second main setting, Calculate Metrics From Forecast Values, to bypass the normality test and always calculate the metrics directly from the forecast data.