Survey research and design in psychology/Tutorials/Multiple linear regression/General steps

Multiple linear regression tutorial - General steps

General steps

edit
View the accompanying screencast: [1]

The general recommended steps for conducting a multiple linear regression analysis are:

  1. Conceptualise the model (e.g., draw a path diagram or Venn diagram to indicate the IVs and the DV) and establish research questions and/or hypotheses.
  2. Check assumptions:
    1. Levels of measurement
    2. Sample size
    3. Normality
    4. Linearity
    5. Homoscedasticity
    6. Multicollinearity
    7. Multivariate outliers
    8. Normality of residuals
  3. Choose the type of MLR:
    1. Standard
    2. Hierarchical
    3. Stepwise / Forward / Backward
  4. Interpret statistical output and psychological meaning of results. Consider:
    1. Overall model
      1. R, R2, Adjusted R2, sig. of R
      2. Change in R2 and the sig. of this change (if a hierarchical MLR is conducted)
    2. Regression coefficients
      1. Y-intercept (labelled "Constant" in the SPSS MLR Coefficients table output)
      2. Unstandardised (B)
      3. Standardised (β (beta))
      4. t and significance for each predictor
      5. Zero-order correlations (r) and semi-partial correlations squared (sr2) for each IV in each model
  5. Depict the relationships in a path diagram or Venn diagram (if useful/relevant)
  6. Regression equation: Present a prediction equation for Y (if useful/relevant)