Topic 1 Statistical Control and Regression Diagnostics

specification errors

occur when the idea or theory a model is based on or the operationalisation of variables is poor or doesnt make sense

total variance

the difference between each individual outcome score and the mean outcome score

explained variance

the difference between each predicted outcome score and the mean outcome score

unexplained variance

the difference between the actual outcome score and the predicted outcome score

unstandardised regression coefficient

the change in the outcome for every one-unit increase in the given predictor, while holding other predictors constant

standardised regression coefficient / beta

the change in the outcome for every one-unit increase in the given predictor, while holding all other predictors constant, when all variables are converted to the same unit of measurement (standard deviation units)

semi partial correlation

the association between the predictor and outcome when the influence of any other predictors is removed from the given predictorsThis allows for the evaluation of the unique contribution of each predictor to the variance in the outcome.

partial correlation

the association between the given predictor and outcome when the influence of all other predictors has been removed from both the given predictor and the outcome

nuisance variables

variables known to influence the outcome but are not of interest to the given model

outliers

datapoints that are unusual based on the score on the outcome variable

influential scores

datapoints that are unusual based on scores on one or more predictor variables

Mahalanobis distance

a measure of influence based on the chi-square value for the number of predictor variables, when p < .001

discrepant scores

datapoints that are unusual based on scores on one or more predictors and the outcome variable

Cooks distance

a measure of discrepancy based on the studentised residuals and leverage scores