XLSTAT: MaxDiff analysis with multiple designs in Excel

 

Background

MaxDiff analysis can be divided in four steps:

1. Choice of the relevant attributes to describe the products.

2. Generation of design of experiments based on incomplete block designs.

3. Collection of the results in Mic<rosoft Excel sheets.

4. Data analysis with a specific method: Logit or Hierarchical Bayes.

 

Dataset for example

A hotel brand wishes to open a new hotel and would like to understand the importance of attributes of a vacation destination in order to define its offer.

The results are divided into different sheets:

1. Attributes: This sheet contains the attributes selected to describe the products.

2. Designs for MaxDiff: This sheet contains the generated design, and the choices given by 5 respondents.

3. MaxDiff: This sheet contains the results of MaxDiff analysis by Hierarchical

 

Step 1: Choosing the attributes

The first step in this MaxDiff analysis is done in collaboration with experts in tourism market. We focus on choosing the important attributes to define a touristic destination. The selected attributes are:

• Quality beaches

• Cultural activities

• Nightlife

• Luxury hotel offer

• Shopping around

• Entertainment

• Wildlife around

• Local cuisine

Among these attributes, we want to evaluate which ones are the most important for tourists. MaxDiff will present sets of attributes to respondents. In every set, respondents are asked to choose which attribute is the most important to them and which one is the least important.

 

Step 2: Generation of the comparisons

XLSTAT allows you to create sets of attributes to be compared by the respondents. A design of experiments method is used.

• Open XLSTAT.

• Click on the CJT icon and choose the Designs for MaxDiff analysis feature.

• Once the button is clicked, the dialog box appears.

In the General tab:

• Enter the Analysis name.

• Select the Attribute names.

• Set the Number of comparisons to 8.

• Set the Number of profiles per comparison to 4.

• Adjust these values to collect enough data for meaningful results while keeping the questionnaire short enough to maintain respondents’ attention.

In the Outputs tab:

• Select the Combination / Respondent format to display combinations in rows and respondents’ responses in columns.

• Set the Number of designs to 2.

• Specify the Number of respondents and Comparisons per design.

• Use multiple designs to increase the variety of combinations presented compared with a single design and reduce potential respondent fatigue or order effects.

• Click on the OK button to launch the computations.

• The results are displayed on a new Excel sheet.

In the Designs for MaxDiff sheet:

• Locate the Choice table.

• Complete the table after collecting the respondents’ choices.

• Enter a value from 1 to 4 for each choice.

• For each respondent, fill in the two corresponding columns:

  • The most important attribute (Best).

  • The least important attribute (Worst).

 

Step 3: Fill in the MaxDiff analysis table

• Fill in the MaxDiff analysis table either manually using responses collected externally or automatically using the individual response sheets. Note: use automatic referencing to simplify data entry, especially when the overall MaxDiff table contains many combinations.

• For each combination, identify the most important and least important attributes selected by each respondent.

• In this example, for Respondent 1 and the first combination, select Entertainment as the most important attribute and Wildlife as the least important.

 

Step 4: Carry out the MaxDiff analysis

• Open the MaxDiff sheet containing the preferences collected from 5 respondents about touristic destinations.

• Make sure the MaxDiff design is filled with the respondents’ responses.

• Click Run the analysis to automatically open the analysis dialog with the data already loaded.

• Alternatively, click the CJT icon and select MaxDiff.

• After clicking on the button, the dialog box appears.

In the General tab:

• Make sure the MaxDiff design was generated with XLSTAT and that the worksheet containing the design has not been manually modified.

• Click the magic wand to automatically load the two required datasets: Responses and Combinations. This will automatically load the two datasets.

• In the pop-up window, select any cell in the worksheet containing the MaxDiff design generated with XLSTAT.

• In this example, select cell B1 in the Design for MaxDiff sheet.

• Click OK to confirm.

Note: you can also select the data manually if needed.

• In the Responses field, select the 10 columns containing the respondents’ responses.

• In the Choices table, select the 4 columns containing the attributes.

In the Options tab:

• Select the Hierarchical Bayes method.

• Leave all other settings at their default values.

• Keep in mind that the Hierarchical Bayes method may take some time to generate results.

• For a simpler and faster analysis, select only Counts analysis in the Outputs tab.

• Click on the OK button to launch the computations.

• The results are displayed on a new Excel sheet.

 

Interpreting the results

The MaxDiff scores are displayed as rescaled coefficients summing up to 100. We see that Individual 1 considers entertainement, local cuisine and quality beaches as very important factors.
Descriptive statistics are also shown. If we check the Means column, we see that Nightlife and Quality beaches are the most important attributes on average. 

We can visualize average MaxDiff scores on the chart below.

After this preliminary study, we may suggest that the hotel chain should focus on luxury hotels with shopping possibilities and cultural activities.