Skip to content

Categorical moderator analysis ​

Open this method

Purpose ​

Assess whether the relationship between X and Y varies with moderator W.

Data and assumptions ​

This release supports one continuous X, one moderator and a continuous Y, with optional covariates. Main effects and interactions enter together. Continuous variables may be centered.

Configure this method ​

W is categorical. Choose its reference group. Interaction coefficients are slope differences from that reference; the conditional-effects table gives each group-specific X slope and interval.

Variables ​

ControlInput
Outcome YSelect one variable
Predictors XSelect variables
Moderator VariableSelect one variable
Control VariablesSelect variables
Treat as categoricalSelect variables
Reference categoriesChoose categorical reference levels

Analysis settings ​

ParameterChoicesDefault
Center continuous variablesSet as neededEnabled
Standard error typeConventional standard errors, HC1 robust standard errors, HC3 robust standard errorsstandard

Analysis parameters remain visible in the main configuration area. Advanced formatting contains only the table title and numeric decimals. Methods that calculate intervals also show the significance level; 0.05 corresponds to 95% intervals.

Read the results ​

Test the interaction before interpreting conditional effects and their confidence intervals. Significance in one group but not another does not itself establish a slope difference. Plotted lines describe the fitted relationship under the stated conditions.

Workflow and exports ​

  1. Upload the dataset and choose the method from the sidebar or search.
  2. Select the variables above, set the design-specific options, then run the analysis.
  3. Inspect the paper-style result. Reconfigure from the result area, or open raw output for diagnostics and sample details.
  4. Rename and keep result tabs as needed. Export Word, Excel, CSV or TXT; batch export can include the selected AI interpretation. Word retains figures. AI text should be checked against the estimates.

Algorithm and references ​