Statistical Sciences 2035 Quiz: Coefficients table terms explained
16 Oct 2018
School
Department
Professor
Document Summary
Uncorrected correlations between y and x = same as in correlations table from the beginning. Look at partial correlation and tolerance: if large tolerance + close to 0 partial- correlations then zero-order correlations are likely to be close to 0. Corrected correlations, that have the overlap between the x-variables filtered out. The square of the semipartial correlation gives the increase of r square if we add this variable to the model: the p-value in the same row shows if that increase of explained variation is significant. B of existing predictor does not change when a new predictor is included, which has no correlation with the other predictor. Standardized values (z-scores) so weights can now be compared. The larger the value, the greater the effect size. If value is exceeded collinearity problem: estimates of regression coefficients would become highly unstable drawing a.
