Multiple Regression Flashcards
What is r?
The multiple correlation coefficient (Persons)
- level of relationship between the model and the DV
What is r2?
Multiple coefficient of determination
- proportion of variance in Y accounted for by X
What is adjusted r2?
for general population
What is the standard error of the estimate?
the typical or average amount by which our predictions will be wrong
- 1-r2 = error variance, the % of variance in Y not explained by X
What is the major difference between correlational and experimental designs?
we cannot infer causation when we have not manipulated our IV (and potentially hold our confounding variables constant). If no experimental design cannot infer causation
What is the bivariate regression equation?
ÿ = a + (b)(X)
What do each of the symbols mean in the regression equation of ÿ = a + (b) (X)
ÿ = the predicted score on the Y variable
a = the y-intercept
b = the slope (B)
X = the persons score on the X variable
What is the z in the standardised regression equation?
z score
What is a residual score?
the difference between the actual score and the predicted score
What is the y-intercept or a?
the constant
What is the difference between the standardised and unstandarised regression equations?
The unstandardised uses B scores
The standardised uses Beta scores
What are beta weights?
are based on common metric and are comparable to each other
- give relative contribution of each predictor (IV) to the overall prediction of the DV
What is a zero order correlation?
- the relationship between the DV and an IV irrespective of any other considerations
- square to get the variance shared by these 2 variables
What is sr (Semi-partial correlation)?
The correlation between what’s left partioning out from X what it shares with Z (part correlation)
- square to get proportion of variance in the DV uniquely explained by the particular IV within a particular model (sr2)
How do you obtain a a sr2 (sr-squared) score?
square part correlation score (in coefficients SPSS output table)
What are the assumptions for multiple regression?
Normality
Linearity
Homoscedasticity
(no) multicollinearity
What is the normality assumption for multiple regression?
residuals are normally distributed
What is the linearity assumption for multiple regression?
the relationships between all variables (between IV’s and DV) are linear
What is the homoscedasticity assumption for multiple regression?
the residuals are evenly spread out around the regression line
What is the (no) multicollinearity assumption for multiple regression?
IV’s are not highly correlated with each other (do not want high correlation as regression coefficients will become unstable)
