-
Variables
-
Explanatory or Predictor
- x-axis
-
Response
- y-axis
-
Predictions
-
Line of Best Fit
-
Least Square Regression
- Line of which the sum of the least squared residuals is smallest
-
Predicted Value
- y-hat
-
Regression to the Mean
- property of the linear model, the line is the linear model
-
Residuals
- Difference between observed and predicted value
- Chart of residuals used to predict linearity
-
Correlation
-
Explains Strength and Direction
- Close to 1 or -1 is strong
- Sign determines positive or negative
-
Measured in r
- r= sum of z scores of x times y divided by n minus 1
-
Warnings
- Doesn't prove linearity
- Only quantitative Variables
- Outliers can distort
-
Not Prove Causation
- Preschoolers have correlation between shoe size and reading level
-
r squared
- measures how successful the regression line is in linearity relating x to y
-
Models
-
Scatter Plots
- A graph to display relationships between 2 quantitative variables measured on same cases
-
Linear Model
-
y-hat=Bsub0 + Bsub1 times x
- Y-hat is prediction
- Bsub0 is y-intercept
-
Bsub1 is slope
- y-units per x-units
- Don't assume linearity, test it out
-
Assessment
-
Direction
-
Positive
- Slope is positive
- bottom-left to upper-right
-
Negative
- Slope is negative
- upper-left to bottom-right
-
Form
-
Decision if Linear
- if it follows a line
-
Scatter
- Is there a pattern?
-
Look for outliers
- y-outlier
- x-outlier
- model outlier
-
Influential Observations
- Change slope of regression line
- If x-value is extraordinary than can use leverage to change slope
-
Influence
- Use Residuals to check for linearity
-
Subsets
-
Look for two separate sets of data
- i.e. boys and girls, age, demographics
-
Bends
- If it bends, it is not linear
-
Extrapolation
- Venturing into x territory not covered by data
- Outliers
-
Lurking Variables
- Variables may be linked by these