This article describes how any outputs saved within Q from advanced analyses can be audited to review how they were created and identify any issues.
Requirements
A Q analysis using advanced methods. The following methods are referenced in the following section:
- MaxDiff
- Choice Modeling
- Latent Class Analysis
- Trees
- Cluster Analysis
- Principal Component Analysis
- Multiple Correspondence Analysis
Method
Segments (latent class analysis and trees)
The settings used to create a Latent Class Analysis or Trees can be accessed by right-clicking on any node and selecting Edit, which identifies the settings used to split the node into sub-segments (most commonly, the same settings are used in all nodes, so there is only a need to right-click on the top node).
The statistical outputs generated when growing the tree are accessed by right-clicking on a segment (i.e., not the top node) and selecting View Report.
Constructed variables
Many of the advanced analysis methods can save variables (e.g., predicted scores from Regression, segments from Cluster Analysis and Latent Class Analysis, factor scores from Principal Components Analysis and Multiple Correspondence Analysis). By right-clicking on these variables in the Variables and Questions tab and selecting Trace Dependencies > Calculated Using you can see the outputs generated when the variables were created.
Experiments
Experiments, such as exotic MaxDiff (anchored, etc) and legacy Choice Modeling, are audited using the tools described in Ad Hoc Audits of Tables (i.e., as from Q’s perspective, such data is a type of question and they are thus audited in the same way as with any other type of question).
Next