IECM 12.0 beta User Manual > Using the IECM > Setting Parameters > Parameter Screens > Uncertainty Editor >

#2: Distribution Menu

The distribution menu is shown on the left, under the parameter information:

The Uncertainty Editor: The Distribution Menu

This menu allows you to choose the distribution to apply to the parameter. The following distributions are available:

•None: The parameter has no uncertainty.

•Lognormal: Lognormal(M, E) describes a skewed (lognormal) distribution where M is the mean and E is the error factor. The standard deviation (s) of the underlying normal distribution is given by s = ln(E) / 1.645. The mean (m) of the underlying normal distribution is given by m = ln(M) ? 0.5s² . The range [ e^m-s ... e^m+s ] encloses about 68% of the probability. The range [ e^m-2s ... e^m+2s ] encloses 95% of the probability, while [ e^m?3s ... e^m+3s ] includes 99%. Note that the error factor does not scale as other parameters do, so the normalized and nominal values will be the same.

•Normal: Normal(m, s) refers to a normal or Gaussian distribution where m is the mean and s is the standard deviation. The range [ m-s ... m+s ] encloses about 68% of this symmetrical bell-shaped distribution. The range [ m-2s ... m+2s ] encloses 95% of the probability, while [ m-3s ... m+3s ] includes 99%.

•Triangular: Triangular(a, b, c) describes a triangular-shaped distribution where the values a, b and c represent the minimum, most likely and maximum values, respectively.

•Uniform: Uniform(a, b) describes a uniform distribution between the deterministic values of a and b. This distribution indicates the uniform probability of a value lying anywhere in the range from a to b.

•Half Normal: HalfNormal(m, s') is a shared distribution; m is the mean and s' is the standard deviation. This distribution reflects the positive part of the normal distribution. It returns the mean value when evaluated deterministically.

•NegHalf Normal: NegHalfNormal(m, s') is a shared distribution; m is the mean and s' is the standard deviation. This distribution reflects the negative part of the normal distribution. It returns the mean value when evaluated deterministically.

•User-defined: UserDefined([ x₀, x₁, ... , x_n ]) allows the user to specify their own samples, bypassing the uncertainty engine. If the number of samples needed is greater than the number of samples specified, the extra samples will all have a value of 1.0 (normalized) or the current deterministic value (nominal). Samples beyond the current sample size may be entered, but they will not be used unless the sample size is increased.