Multiple Regression Lab Chapter 10 1. Topics Multiple Linear Regression Effects Levels of...

Post on 02-Jan-2016

220 views 0 download

Tags:

Transcript of Multiple Regression Lab Chapter 10 1. Topics Multiple Linear Regression Effects Levels of...

Multiple Regression

Lab Chapter 10

1

Topics

• Multiple Linear Regression• Effects• Levels of Measurement• Dummy Variables

2

Multiple Linear Regression

3

Output from Regressing

INCOME86 on EDUC, AGE, and

SEX

Model Summary• Multiple correlation coefficient R– Correlation of all IV’s with DV– Report strength

• Coefficient of determination Adj. R2

– % of explained variability in DV by all IV’s

ANOVA Table- Size of F and p-value (sig.) indicate significance of overall model- Large F, small p-value (< .05 or .01) is a significant model

Coefficients Table• Similar to bivariate regression– unstandardized coefficients– regression equation

Y = -8142.153 + 1650.831(X1) + 267.101(X2) - 7174.415(X3) INCOME86 = -8142.153 + 1650.831(EDUC) + 267.101(AGE) - 7174.415(SEX)

Multiple Linear Regression (cont.)

• standardized coefficients (betas)– net effects– indicate direction and strength

8

Net Effects

• Interpretation– “What effect does this variable have separate

from the effects of the other independent variables?”

– a way to statistically control for other variables in the model

• A potential problem: multicollinearity (more on this in lecture)

9

Levels of Measurement

Can it be used in linear regression as

Dependent variable Independent variable

Interval/ratio variable yes yes

Ordinal variable Yes (but with caution) yes

Dichotomy no yes

Nominal variable (with three or more attributes)

no Maybe (as series of dummy variables)

10

Dummy Variables• a way of getting nominal variables with 3 or

more attributes into regression as independent variables

• conversion into a series of dichotomies• enter all but one of the dichotomies

Original Variable REGION Northeast = 1, Midwest = 2, South = 3, West = 4 New Variables NOREAST if REGION = 1, NOREAST = 1 otherwise NOREAST = 0 MIDWEST if REGION = 2, MIDWEST = 1 otherwise MIDWEST = 0 SOUTH if REGION = 3, SOUTH = 1, otherwise SOUTH = 0 WEST if REGION = 4, WEST = 1, otherwise WEST = 0