Marketing - MaxDiff - Ensemble

From Q
Jump to: navigation, search

Create an ensemble of MaxDiff models by averaging the respondent parameters. The output may either be a table comparing the underlying models and the ensemble, or histograms of respondent parameters of the ensemble. In either case, MaxDiff functions to save variables (e.g. Individual-level Coefficients) may be applied to the ensemble.

Example

To run the ensemble in Displayr, use the Compare Models function by selecting Insert > More > Marketing > MaxDiff > Compare Models and select the Ensemble option.
To run the ensemble in Q, select Create > Marketing > MaxDiff > Compare Models and select the Ensemble option.

Options


MaxDiff models created in Displayr and Q are used as the inputs to the ensemble. The inputs are shown in the image below.

Ensemble inputs.PNG

Options

Input models At least 2 MaxDiff models.

Ensemble Whether to create an ensemble by taking the average of the respondent parameters across the models.

Output

Comparison A table comparing metrics from models (and the ensemble, if selected).
Ensemble Histograms of respondent parameters, as per the underlying MaxDiff outputs.

Histograms of ensemble parameters

More information

See Comparing MaxDiff Models amd Creating Ensembles blog post for an example of comparing models.

Code

var controls = [];
var modelsInput = form.dropBox({label: "Input models", types:["RItem:FitMaxDiff"], name: "formModels",
                               multi: true, required: true, min_inputs: 2,
                               prompt: "Select at least 2 MaxDiff models."});
controls.push(modelsInput);

var ensemble = form.checkBox({label: "Ensemble", name: "formEnsemble", default_value: true,
                              prompt: "Whether to create an ensemble of the models."});
controls.push(ensemble);

if (ensemble.getValue()) {
    var output = form.comboBox({label: "Output", 
              alternatives: ["Comparison", "Ensemble"], name: "formOutput", default_value: "Comparison",
              prompt: "A table comparing the models, or histograms of ensemble respondent coefficients."});
    controls.push(output);
}
form.setInputControls(controls);

if (ensemble.getValue())
    form.setHeading('Ensemble of MaxDiff Models');
else
    form.setHeading('Compare MaxDiff Models');
library(flipMaxDiff)
ensemble <- MaxDiffEnsemble(formModels,
                              !formEnsemble,
                              get0("formOutput", ifnotfound = "Comparison"))

Further reading: MaxDiff software