POLI 210 Lecture Notes - Lecture 18: Null Hypothesis, Central Limit Theorem, Statistical Process Control
Document Summary
Minimize the errors: the sum of squared residuals. Residuals are vertical deviations from the line. The regression coefficient: condition for calculating. : a one-unit increase in x produces a unit increase in unit increase in y. When x is zero, the predicted value for y is. The regression line always passes through two points: This may or may not be meaningful: point, point. Substantive significance: the size of the effects. Causal significance: the plausibility of the model. A residual plot is a scatterplot of the regression residuals against the explanatory variable x or the predicted values. The residual plot is a diagnostic plot: we search for signs of patterns. Ideal residuals should look like the outcome of pure chance. Interpretation: proportion of the total variance explained by the fitted model. We need to assume that all other factors affecting y are random and uncorrelated with: this is often unrealistic.