Using Ngene for advanced DCE design
NGENE is required for advanced DCE designs - Bayesian priors, complex constraints, generic-specific parameters, and designs SurveyEngine's built-in generator does not support.
What NGENE is, when to use it instead of SurveyEngine's built-in generator, and how to import NGENE designs into SurveyEngine.
Knowledge Base -> Experiment Design -> Methods & Academic
Ben White, 07.07.2026
When the SurveyEngine design generator is not enough
SurveyEngine's built-in generator handles standard D-efficient designs. NGENE is needed when the design problem is more complex.
NGENE supports Bayesian D-efficiency with informative priors, generic-specific attribute constraints, nested and cross-nested designs, non-linear utility specifications, and complex level balance constraints.
Bayesian D-efficiency and priors
Bayesian D-efficiency uses prior parameter information - from a pilot study or the literature - to produce designs efficient at the expected parameter values. For studies with prior information, Bayesian designs can reduce required sample sizes by 30-50%.
TLDR Quick links
Importing Ngene designs into SurveyEngine
Step 1: Get Access to Ngene Online from Choice Metrics
Step 2: In your Ngene™ account, write the design file specifying attributes, levels, utility functions, and design objectives. The Ngene™ manual covers the syntax.
Step 3: Run the design. Ngene™ generates the design matrix and reports D-efficiency. Iterate until the efficiency target is achieved.
Step 4: Export the design matrix as a CSV with one row per choice set and columns for each alternative's attribute levels.
Step 5: Import into SurveyEngine using the design import function. The system maps design columns to your choice experiment attributes.
Step 6: Preview the choice tasks to confirm attribute levels display correctly for each choice set.
Worked example - Bayesian design for transport study
A transport mode choice study requires a generic-specific design with car-specific attributes (fuel cost, parking cost) and rail-specific attributes (seat availability, punctuality), plus shared attributes (travel time, frequency). SurveyEngine's generator does not support generic-specific constraints.
The design is specified in Ngene™ with Bayesian priors from a previous mode choice study as the Built-In Ngene Lite is runtime limited to 60 seconds. Ngene™ produces a 12-task design with 94% Bayesian D-efficiency. The design matrix is exported from Ngene™ and imported into SurveyEngine. The Choice tasks display correctly across all alternative configurations.
References
Bliemer, M.C.J. and Rose, J.M. (2010). Construction of experimental designs for mixed logit models allowing for correlation across choice observations. Transportation Research Part B, 44(6), 720–734.
Ready to generate a Bayesian D-efficient design? Log in to SurveyEngine and open the Ngene script builder.
Or Contact us at support@surveyengine.com — we're glad to help.