Modeling

Inherent Uncertainty

Inherent or physical uncertainty is associated with inherent random nature of a measurable process’ variable. Examples include the variation in material strength, variability of wind loads, dimensions of a structural component, time to complete a task, price per square foot of residential property in a neighborhood, speed of a car, etc. The inherent uncertainty for any basic variable must be estimated from observations of the variable or be subjectively assessed.

Statistical Uncertainty

Statistical uncertainty arises from the inaccuracies in estimation of the statistical parameters of a random variable and in choice of the probability distribution. For example, when the available data are limited, the estimated mean and variance of a variable may not be accurate and the probability density function determined on the basis of available data may not be the most appropriate. This source of uncertainty arises from the limited nature of the available data. A small data size may result in uncertainty in the estimated model parameters, even if the model form is correct and the measurements are exact. Uncertainties associated with the availability of statistical data can be reduced by increasing the amount of data.

Model Uncertainty

Model uncertainty is caused by the use of a simplified relationship between the variables to represent the real relationship or phenomenon of interest. Model uncertainty has two components: one due to lack of understanding of the phenomenon itself, and the other due to the use of simplified models. In its simplest form, modeling uncertainties concerns the uncertainty of predictive models, such as a mathematical expression that estimates the value of a house based on square footage of the house living space, price per square foot of the living space, square footage of the land, price per square foot of the land, number of bedrooms, number of bathrooms, etc.

Measurement Errors

Predictive models are constructed, assessed, verified, validated, and updated based on measured values of the observable model variables. Often these values are in error due to inaccuracies in measurement procedures and/or devices. A model constructed, assessed, verified, validated, and updated on the basis of such erroneous data will clearly contain uncertainty.

Human Error

Uncertainty due to human error arises from errors made by analysts, designers, or operators in the design, construction, or operation phases of a system. Examples may include calculation errors or omissions in the design phase of a structure, error in the placement of rebar in reinforced concrete construction, and errors in the operation of a structure which result in its exposure to overloads.

Other Types of Uncertainty

Other types of uncertainties that might be considered in probabilistic analysis include uncertainty due to human intervention, prediction uncertainties, phenomenological uncertainty, and decision uncertainty.