MaxDiff is a way to understand the strengths and weaknesses of an evaluation study. It is an approach to evaluate and describe an evaluation, not an approach to develop a questionnaire or collect data.
Maxdiff is primarily used to evaluate preferences, but it has been used as well to compare observed behavior, such as capturing the percentage of people by the time of day who are salmon fishing in a particular river.
Use of MaxDiff:
The purpose of this guide is to help implement a MaxDiff analysis. This is a method for evaluating objective data such as environmental conditions and health outcomes for policy analysis, evaluation, and research.
MaxDiff Analysis should be applied to all quantitative research to produce objective data for decision-makers and policymakers. A MaxDiff analysis uses integrated metrics that combine multiple dimensions of human health and environment responses to provide insights about causes or effects on various types of mortality (e.g., life expectancy) from environmental changes or other processes such as “natural” disasters or population trends.
The focus on integrated metrics is important because the different types of mortality are often very closely related but can be difficult to combine into a single number or metric.
The methodology of MaxDiff was designed to overcome the general challenges in data collection, analysis, and understanding the validity of data sources. The research approach should not be limited by the use of specific software, rather a consistent procedure should be used with any appropriate statistical software.
Steps To Implement MaxDiff Analysis Effectively:
At each step of the process, an objective review should take place to ensure that any subjective judgments do not lead to bias in the results.
The basic steps are:
1) Identify the problem or issue that you are researching.
Doing this will help to determine the type of analysis you will use.
2) Choose your population variable.
This step is primarily based on your problem or issue. The general population or a specific sub-group. Be sure you have an appropriate measure of exposure. Do not let this be limited by the availability of data. Look at multiple sources and try to find more than one measure of exposure to get an idea of the level and variation in this measurement.
If you can think up different exposure options, do so and simulate that data with MaxDiff software to see if they are reasonable in order to increase your likelihood of getting valid answers.
3) Design the questionnaire to use for data collection.
Make sure that the questionnaire is designed with at least four (4) variables or dimensions. You will need to use the MaxDiff software to help determine if your design meets the requirements of your problem or issue.
4) Test your questionnaire
Do this with at least five people to ensure that it is working properly and is understandable. Then, analyze your data. It is important to collect the data yourself if possible to increase the validity of your findings.
In some instances, you may need to use a survey firm or third party to administer and code your questions. If this is the case, you will want to check with this company prior to using them to ensure that they have an understanding of MaxDiff Analysis and will be able to produce results using this methodology consistent with what you expect.
5) Assign levels, or categories, of responses using a standard scale (e.g., Likert scale).
This is to ensure the best quality distribution of data.
6) Prepare instructions
This is for the participants before conducting data collection, including process instructions and compensation instructions if any. This will have to be translated often into the language of the participants.
7) Collect The Results
Conduct data collection by asking participants to rank items according to preference on their personal level; then write down the results in descending order of preference.
In this way, you’ll be able to implement it most effectively and help your organization or company make the best decision possible for which you’ll have to create tables, charts, or graphs of the data to see any patterns and help determine causal relationships between the different variables.