Limits Overview

The LowLow, Low, High and HighHigh properties for all continuous values can be defined in the preparation step. Any values that fall outside the LowLow and HighHigh limits specified for each individual process variable are considered to be of bad quality and will be disregarded in further steps. It is good practice to set these limits to exclude any values that are outside the engineering limits of that particular field. Limits can also be used to eliminate outlying data points from the data set.

When loading a dataset into the Discrete & Batch Troubleshooter, limits are automatically calculated for each variable in the data range, and disabled by default. The limits are disabled as a default when loading a dataset for the first time to prevent the following situation: When switching datasets to load for process modeling, if there is a big difference in the data range between the two datasets, the new data could fall completely out of the limit range of the initial dataset. This will cause all the new data to be marked as bad quality. The simplest way to re-define these limits for any variable is to simply enter the appropriate limits manually.

There are four limits that can be configured for each variable:

Low Low

The Lowlow limit is the absolute (engineering) minimum value that a model variable may have. A typical use of the LowLow limit is to set the data quality of all values that are less than the LowLow limit to "bad quality". All "bad quality" data will not be used in model construction.
Low

The Low limit is a secondary limit that indicates the "operational" lower limit of a variable. By default, the low limit is set to 1 standard deviation from the mean.
High

The High limit is a secondary limit that indicates the "operational" higher limit of a variable. By default, the high limit is set to 1 standard deviation from the mean.
High High

The HighHigh limit is the absolute (engineering) maximum value that a model variable may have. A typical use of the HighHigh limit is to set the data quality of all values that are less than the HighHigh limit to "bad quality". All "bad quality" data will not be used in model construction.

Use the following visualization techniques to identify ranges and outlying data points in your data set: