What is Correspondence Analysis
Correspondence Analysis (CA) is a powerful method used to explore relationships between a multitude of categorical variables. It transforms multifaceted data into a simple graphical representation, making it easier to uncover patterns and associations within datasets.
To analyze objects, CA first reduces the dimensionality of the initial feature space. This is done by identifying a certain n-dimensional space that best represents the original data with minimal loss of information.
Next, the obtained axes (factors) are rotated to distribute the overall variance among all the discovered factors. The goal is for the first factor to explain most of the variation between objects, the second factor to explain less, the third factor to explain even less, and so on.
As an outcome, the objects are distributed on a 2-axis plot, enabling researchers to visually explore the relationships between features and objects, as well as between objects themselves.
How CA is used in market research
Correspondence analysis is a valuable tool for the identification of key patterns, enabling more informed decision-making and marketing strategies. Many industries use it to yank actionable insights.
The most common application of CA in market research is brand position. Here brands of a particular category are the objects, while brand attributes assessed in a survey are the features. The CA result is a brand image map used to represent the branding and positioning of the different brands in the market.
This method can be used for:
customer segmentation;
identifying the most differentiating features, developing a unique selling point;
social media listening (SML) analysis;
prioritizing customer satisfaction factors;
assessing the impact of advertising campaigns on the market;
prioritizing product features.
The process of Correspondence Analysis in DataTile
Correspondence Analysis service must be activated on the server before it can be launched.
Step 1. Select variables
Start with creating a contingency table containing objects (fictional periodical press brands in the example) and descriptors as rows (emotions people feel about each brand).
You can use any metric at this stage. The table should contain at least 4 rows and 4 columns.
Step 2. Set-off the analysis
Launch the Correspondence Analysis module by clicking the 'download' icon in the upper right corner.
This will create a new item of a specific type in the My Space folder. You can always access the results of your analysis by opening this item.
Step 3. Check the model quality
Before interpreting the results check whether the statistical criteria for the model are met. You can see the summary information by clicking the 'info' icon in the upper right corner when the CA module is opened.
No further analysis will be run if the model quality is too low. Please return to the previous step and check your data.
Step4. Interpret the results and adjust the visualization
If the model quality meets the standard, you will see a set of tabs showing the results of the CA.
Use the Axis interpretation tab to select which factors to use in the final solution. Most likely you will want to use no more than two. Here you can interpret the axes based on the coordinates, inertia or cosine indicators for each dot. After you’ve chosen your axes you can give them titles, which will be shown on the diagram.
Use the Diagram tab to adjust the visualization of the selected axes. You can hide and show dots, change labels, and limit the number of dots shown on the scatter plot.
The Residuals tab can be used to better access the relationship between column and row points on the chart. It is strongly recommended to avoid interpreting the closeness of those types of dots as their relationship. You can see the same data if you hover over dots on the scatter plot. Please read more about the standardized residual analysis here.
Step5. Save the results
To export the result of CA into a PowerPoint slide click the red button in the upper right corner.
