Dimensions and patterns

On the Select Criteria page, you can specify the dimension and pattern for an MGPS run as criteria for viewing data mining results. To display these controls, click Show Advanced.

Dimensions

The dimensions in a run are the number of ways in which items combine to generate scores. When specifying data mining parameters for MGPS runs, the run creator specifies the highest dimension to compute. The run results include combinations of all dimension combinations, from 1 to the specified highest dimension. For example, a two-dimensional (2D) data mining run performed for drug and event item variables produces scores for the following combinations:

Drug (1D)

Event (1D)

Drug+Drug (2D)

Drug+Event (2D)

Event+Event (2D)

where

Drug and Event represent the names of the item variables used for the data mining run.

Note:

One-dimensional results are useful to determine the total count (N) for a drug or event. Other statistics are not generated for 1D results, and you cannot display graphs for 1D results.

A 3D run for the same drug and event item variables produces results for these combinations:

Drug (1D)

Event (1D)

Drug+Drug (2D)

Drug+Event (2D)

Event+Event (2D)

Drug+Drug+Drug (3D)

Drug+Drug+Event (3D)

Drug+Event+Event (3D)

Event+Event+Event (3D)

Note:

If you include PRR, RGPS, or ROR computations in an MGPS run, Oracle Empirica Signal produces PRR, RGPS, and ROR scores only for Drug+Event (2D) combinations.

When specifying criteria to view data mining results, you can indicate the number of dimensions to view. You can include 1 dimension up to the highest dimension included in the run. To include all dimensions, select All. The default is the highest dimension included in the run. If you change the default number of run dimensions, consider changing the default pattern for the run.

Note:

Combinations of two or more dimensions that include only multiple drugs or only multiple events (such as Drug+Drug or Event+Event) are only included in the results table if the run creator specified (on the Data Mining Parameters page) that results for items of the same type should be kept.

Patterns

You select a pattern to define the item types (such as drugs and events) to appear in the results table or graph. The default pattern depends on the number of dimensions in the run. The following table shows the default pattern for each dimension. Drug and Event represent the names of the item variables used for the data mining run.

Dimensions	Default Pattern
1D	Drug
2D	Drug+Event
3D	Drug+Any Drug+Event

The results table or graph includes results only for the selected pattern. You can change the types of items supplied by the default pattern as needed. The following table describes the options for the pattern fields:

Option	Description
<variable-name>	Lists each item variable used in the data mining run; for example, Generic_Name and PT. (Item variables can also represent other types of data, such as Gender or Age_Group.) For any pattern that includes <variable-name>, you can specify terms for the variable. If you specify terms, the results table displays only those terms and the scores calculated for them; otherwise, the results table displays all values for the <variable-name>.
Any <variable-name>	Lists each of the item variables used in the data mining run with the prefix Any; for example, Any Trade_Name and Any PT. These pattern options allow you to display results for specified terms as though you had supplied an OR between those terms. For example, if you choose the pattern PT+PT and specify the terms Headache and Rash, the results table displays only rows that contain both Headache and Rash. However, if you choose the pattern PT+Any PT, then specify Headache and Rash, the results table displays rows that contain either Headache or Rash.
Any Item	The Any Item option is a wildcard that represents any of the other item variables in the run. You can use Any Item in a pattern for a 3D run that uses the Generic_Name and PT item variables. When you choose the pattern Any Generic_Name+Any Item+PT and specify the term DrugA, the results table displays DrugA+Event+Event combinations and also DrugA+Drug+Event combinations.