Vocab Unit 6

Subsets

Subsets

Groups within a larger pool of data that are different in a specific way and thus often should be considered separately when creating linear models

Extrapolation

Using a linear model to predict values beyond those found within the domain of the data… can be highly unreliable, as it assumes that all other conditions still hold

Influential Point

A point that results in a very different slope for a regression model if it is removed

Outlier

A y-value that is far from its predicted value, resulting in a large initial residual… may or may not be an influential point

Leverage

A data value whose x-value is far from the mean of x has high leverage, like an outlier in the x direction rather than y… may or may not be an influential point

Lurking Variable

A hidden variable that simultaneously affects both variables in an association, accounting for the correlation that may appear between the two

Residuals Plot

A scatterplot of the residuals versus the x-values of an association, with the x-axis denoting a residual of 0… a residuals plot with no apparent pattern (blob), bouncing above and below the x-axis, means that the determined LSR model is appropriate

Re-Expression

Transforming a data set by taking the logarithm, square root, reciprocal, or some other math operation of ALL values in the data set to make it more conducive for linear regression

Nearly Normal Residuals Condition

To perform inference for regression, the residuals must be ~Normally distributed [linear normal probability plot]

Straight Enough Condition

To perform inference for regression, the association (scatterplot) studied must be ~linear [check residual plot]

Equal Variance Condition

To perform inference for regression, the variability of y must be ~constant for all values of x; check the spread of the residuals around the predicted value of the residuals plot

Standard Error of the Slope

The variation of the slope due to sampling variability, which is influenced by three factors: spread about the line (se), spread of the x-values (sx), and sample size (n)