Partial and Semi-Partial Correlation Flashcards
• Why would we want to compute partial and semi-partial correlations?
o We compute partial correlations when we want to hold 1 variable constant over the other (“over and above the effect of…”
• Be able to define partial and semi-partial correlation (in terms of correlations among residualized variables).
o Partial correlations is when we remove the shared variability of Z-Y when finding X and Y correlation.
• Correlation of X and Y while holding Z constant
o Semi-partial looks at the unique amount of overlap bw X and Y where Z has been removed from X but not Y. That second predictor is still affecting the outcome variable.
• Correlation between X and Y over and above the effect of Z. All of Z is removed, except for what’s shared with Y is removed from the X and Y correlation.
• Be able to recognize and produce notation for partial and semi-partial correlations.
o Partial – ryx.z (ryx controlling for z)
o Semi-partial – ry(x.z) (z is removed from x but not y)
• Be able to interpret partial and semi-partial correlations.
o Partial – This is the relation between X and Y when the effects of Z have been removed from BOTH (by holding Z constant)
o Semi-partial – The proportion of variance in Y accounted for by X and over and above Z
• Know which correlation is larger (further from zero) and why.
o Zero-order – It shows shared variability ignoring everything else
o Partial is second biggest because the amount of Y is smaller; total amt. of variability in Y is smaller, so the shared portion takes more portion of Y.
o Semi-partial is smaller than both