How Are Exceptions Evaluated? :: SAS(R) IT Resource Management 3.7: Administrator's Guide

Overview of the Exception Evaluation Process

Exceptions are identified by evaluating the data in the input table, which can be filtered, using the criteria specified in the exception definitions that are listed for the transformation. Exceptions that are identified are written to the target exception table. (You can also specify that more detailed information be written to the exception condition table). For information about how to specify an exception, see Working with the Exception Transformation.

Filtering Data for the Exception Evaluation Process

The data is first filtered according to any specifications in the Exception transformation’s Filters tab. The filter can be based on a user-specified expression such as a list of resources that you are interested in. The following expression is an example of this type of filter:

MACHINE IN (‘one’, ‘two’, ‘three’)

The filter can also specify other user requirements such as filtering by router type, or excluding time periods that are not of interest. The following expression is an example of this type of filter:

DAYOFWEEK^=’Sat’ AND DAYOFWEEK^=’Sun’

The Last Time Periods field can also be used to filter the incoming data.

If a filter is not defined, then all the input data is considered for evaluation.

Evaluating the Exception Definition

After the data is filtered, each observation is evaluated using the exception expression from each of the exception definitions in the transformation. For each incoming observation (of the filtered data), the exception expression is determined to be either true or false for each exception definition.

At this point, observations for which the expression is true are only potential occurrences of an exception. This is because there are grouping and occurrences attributes of the exception definition that factor into whether an exception is found or not. These values affect whether an exception has been detected:

Grouping is specified on the Specify Grouping page of the Exception Definition wizard.

Group By columns define the grouping that organizes the data to evaluate. After the grouping data is evaluated, a set of potential exceptions is identified. If no grouping is specified, then the entire incoming data is evaluated as a single group.

Note: In addition, if no Group By column is selected, “For All Data” is displayed on the overview report, rather than a specific by group values, if that report is generated.
The occurrences attribute of the exception definition is used to evaluate any potential exceptions to determine whether actual exceptions exist. The occurrences value is specified on the top part of the Occurrences and Exception Type page of the Exception Definition wizard. The occurrences part of the page is outlined in red in the following display:

Occurrences Part of the Occurrences and Exception Type Page of the Exception Definition Wizard

Processing Exceptions

An exception is flagged only if both the exception expression and the occurrences specification are satisfied.

If the criteria of an exception definition are met, one observation is written to the exception table, which is considered to be one exception. At the same time, for that group, all observations for which the expression is true are written to the exception condition table. (If you have not specified an exception condition table, this information is written to an internal work table.) If no exception is found for a given group, then nothing is written for the group to either table, and nothing further is done for that group.

For example, suppose that the expression was true for 5 observations in a group for a particular exception definition. However, the Occurrences section of the exception definition specified a Percentage of observations in the selected group of 25%. If a group consisted of 25 observations, and the expression was true for 5 observations (which is 20% of the group), the criteria for an exception are not met. Therefore, no exception would be found. Because no exception was found, the following statements are applicable:

No observation is written to the exception table.
No observations are written to the exception condition table, if the creation of this table was specified.

However, using the same example, imagine if the expression was true for 7 observations. In this case, an exception would be found, because 28% is above 25%. Because an exception was found, the following statements are applicable:

One observation is written to the exception table.
Seven observations are written to the exception condition table, if it was specified.

Handling Missing Values

Missing Values in the Constant Threshold, Other, Range, and Statistic Bounds Types

If an observation has missing values for any numeric columns that are specified in the expression, then the observation is skipped before the data is evaluated. (This does not apply to character columns with missing or blank values.)

For example, suppose the exception definition has these characteristics:

Occurrences: 3 in a row
Type: Constant
Analysis column: AVENQUEMAX
Expression: AVENQUEMAX > 1

If the incoming observations have the pattern of missing values that is shown in the following table, then an exception is found. The exception is found because the observation at 01Jan2013:00:30 is omitted before evaluation (due to a missing value for AVENQUEMAX). Thus, three observations in a row are detected where AVENQUEMAX>1.

Table of Incoming Observations
Datetime	AVENQUEMAX
01Jan2013:00:00	1.234
01Jan2013:00:15	1.329
01Jan2013:00:30	.
01Jan2013:00:45	1.002

Note: A value of ‘.’ for a numeric column is the standard SAS way of denoting a missing value.)

Missing Values in the Free-Form Type

For free-form exception definitions, there is no automatic detection of missing values. Because you supply the complete form of the expression, you are also responsible for whatever treatment of missing values that is appropriate to the task.