4.1 Data Types
Data fall into one of three types:
- Dichotomous data can have one of only two values. In the process industry examples might include pass/fail status of a check against a product specification. Similarly it might be the pass/fail status from validating a measurement from an on‐stream analyser or inferential. Such data can be averaged by assigning a numerical value. For example, we might assign 1 if a MPC application is in use and 0 if not. Averaging the data would then give the service factor of the application.
- Nominal data have two or more categories but cannot be ranked sensibly. For example, the oil industry produces multiple grades of many products. Specifications can vary by application, by season and by latitude. For example, a data point might have the value ‘summer grade European regular gasoline’. Only limited mathematical manipulation is possible with such data. For example, it would be possible to determine what percentage of cargoes fell into this category.
- Cardinal data have two or more categories that can be ranked. Most process data fall into this category. Within this type, data can be continuous or discrete. Process measurements might generally be considered as continuous measurements. Strictly a DCS can only generate discrete values although, for statistical purposes, the resolution is usually such that they can be treated as continuous. Laboratory results are usually discrete values. This arises because testing standards, ...