v0.4.3
Wiki/Statistical Procedures

Linear Regression & Diagnostics

Ordinary Least Squares (OLS), ANOVA partition, collinearity diagnostics (VIF, Tolerance), and Durbin-Watson autocorrelation.

Verified vs SciPy (≤ 10⁻⁶)IBM SPSS Algorithms pp. 883–894

1. OLS Formulation & Normal Equations#

Linear regression solves the coefficient vector b\mathbf{b} via QR/Cholesky decomposition of the normal equations:

b=(XTX)−1XTy\mathbf{b} = (\mathbf{X}^T \mathbf{X})^{-1} \mathbf{X}^T \mathbf{y}

StatCraft outputs full standardized coefficients βj\beta_j, standard errors, tt-values, and two-tailed significance levels.

2. Collinearity Statistics (VIF & Tolerance)#

For each predictor jj, the tolerance Tolj=1−Rj2Tol_j = 1 - R_j^2 and Variance Inflation Factor VIFj=1/ToljVIF_j = 1 / Tol_j are computed to detect multicollinearity.

3. Residual Autocorrelation#

Durbin-Watson dd is computed from residual differences to verify independent errors:

d=∑i=2N(ei−ei−1)2∑i=1Nei2d = \frac{\sum_{i=2}^N (e_i - e_{i-1})^2}{\sum_{i=1}^N e_i^2}

Canonical SPSS Syntax Example

SPSS SYNTAX (.SPS)
1REGRESSION
2 /DEPENDENT salary
3 /METHOD=ENTER educ prevexp
4 /STATISTICS=COEFF R ANOVA COLLIN TOL ZPP.