Description
Fitting multiple linear regression to UX outcomes while formally checking linearity, independence, homoscedasticity, normality of residuals, multicollinearity and influential cases.
Multiple regression estimates how predictors such as task success, perceived ease and tenure relate to an outcome such as satisfaction, holding others constant. Its estimates and p-values are only trustworthy if assumptions hold: linear relationships, independent errors, constant error variance, approximately normal residuals (for small-sample inference) and no excessive multicollinearity, with no single cases driving results. SPSS REGRESSION provides residual plots, Durbin-Watson, VIF and tolerance, Cook's distance and leverage. Rigour comes from a pre-specified model, inspecting diagnostics rather than just R squared, using robust standard errors or transformations where needed, and not interpreting regression coefficients causally without a causal design.
Attributes
- Discipline
- Statistics & Modelling
- Loss type
- Vestigial
- Stage
- Evaluation, Post-launch
- Data
- Quantitative
- Effort
- Days
Steps
- Specify the model from theory, including a sample-size check (common heuristic: at least 10-20 cases per predictor).
- Run REGRESSION from syntax with /STATISTICS COEFF CI R ANOVA COLLIN TOL and /RESIDUALS DURBIN, saving ZRESID, ZPRED, COOK and LEVER.
- Plot standardised residuals against predicted values to check linearity and homoscedasticity, and a Q-Q or histogram of residuals.
- Check VIF (investigate above about 5, act above 10) and Durbin-Watson (roughly 1.5 to 2.5).
- Inspect cases with Cook's distance above 1 or above 4/n, and standardised residuals beyond plus or minus 3, re-running with and without them.
- Use HC3 robust standard errors or transformations if heteroscedasticity is present, and report standardised and unstandardised coefficients with CIs.
Sources
- SPSS REGRESSION
- PLUM
- LOGISTIC REGRESSION procedures; Hayes PROCESS macro v4
- Field, A. (2024) Discovering Statistics Using IBM SPSS Statistics, 6th ed.
- Cohen, J., Cohen, P., West, S. G. & Aiken, L. S. (2003) Applied Multiple Regression/Correlation Analysis for the Behavioral Sciences, 3rd ed.
- The Inquiry Series, C8 — SPSS for Regression, Moderation & Mediation
Also known as
- OLS regression diagnostics
- Multiple linear regression
- Variance inflation factor (VIF) check
Categories
- Analytics & Behaviour
- Statistics & Modelling