content format

How to Run SI-CHAID in Statistical Software SI-CHAID is a specialized statistical software package developed by Statistical Innovations. It expands upon the traditional Chi-squared Automatic Interaction Detection (CHAID) algorithm. The software excels at building segmentation trees, profiling target groups, and handling large datasets with missing data or complex survey weights.

Running an analysis in SI-CHAID requires a structured approach to prepare your dataset, configure the tree parameters, and interpret the visual results. Step 1: Prepare and Import Your Data

Before opening the software, format your dataset to ensure a clean import.

Format the file: Save your dataset as a compatible file format, such as SPSS (.sav), Excel (.xlsx), or delimited text (.csv, .txt).

Define variable types: Ensure your target (dependent) variable and predictor (independent) variables are clearly defined as nominal, ordinal, or continuous.

Handle missing values: SI-CHAID natively handles missing data by treating “missing” as a separate valid category or floating category, so you do not need to delete these rows.

Launch and import: Open SI-CHAID, click File > Open, select your file type, and load your dataset into the workspace. Step 2: Configure the Analysis Model

Once your data is visible in the spreadsheet viewer, you must define the roles of your variables.

Select the dependent variable: Click on the variable you want to predict or segment, right-click, and designate it as the Dependent (target) variable.

Select independent variables: Highlight the columns you want to use as predictors, right-click, and designate them as Independent variables.

Assign weights (optional): If you are analyzing stratified survey data, select your weight variable and assign it as the Case Weight to ensure accurate population estimates. Step 3: Adjust the CHAID Parameters

Tailor the tree-building algorithm to match your sample size and research goals by modifying the project specifications.

Set significance levels: Navigate to the analysis options to set the alpha (

) threshold for splitting and merging nodes. The default is typically

Adjust Bonferroni correction: Keep this enabled to control for Type I error inflation caused by making multiple statistical comparisons across categories.

Define node sizes: Set the minimum number of observations required for a parent node to split (e.g., 100 cases) and the minimum size allowed for a child node (e.g., 50 cases) to prevent overfitting.

Choose the split method: Select between standard CHAID (which merges categories first, then splits) or Exhaustive CHAID (which looks at all possible splits, providing a more thorough but computationally intensive search). Step 4: Run the Algorithm and Explore the Tree

With your variables assigned and parameters locked in, you are ready to generate the segmentation model.

Execute the run: Click the Run button (often represented by a green arrow or gears icon) on the main toolbar.

Review the tree diagram: The software will automatically generate an interactive, color-coded tree diagram showing the optimal splits.

Inspect the nodes: Look at each node to see the sample size (

), percentage breakdown of the target variable, and the Chi-square statistic ( X2cap X squared ) along with its corresponding

Prune or grow manually: If a split does not make practical sense, right-click the node to manually force a split or prune the branch back to simplify the model. Step 5: Export the Results and Model Rules

The final step is to save your insights and apply the model rules to new data.

Save the visual tree: Go to File > Export to save the tree diagram as an image file (PNG or JPEG) or a PDF for presentations.

Export syntax and rules: Export the scoring logic as SPSS syntax, SAS code, or SQL statements to easily score new databases outside of the software.

Save the project: Save your workspace as an SI-CHAID project file (.chaid) so you can modify your parameters or update the data later.

To help me tailor this guide further, could you share a bit more about your project? Which file format is your dataset currently saved in?

What is your dependent variable (e.g., customer churn, test scores)? Do you plan to use standard CHAID or Exhaustive CHAID?

Knowing these details will allow me to provide specific syntax examples or step-by-step optimization tips.