SPSS Chapter 9 Dummy Variables and Interaction Effects

1
SPSS Chapter 9Dummy Variables and Interaction
Effects

• PS 371

2
Review

• Regression is about predicting changes in the
  dependent variable based on changes in the
  independent variable(s)
• This is a bivariate linear relationship
• Y a bx1 e

3
Multivariate analysis

• Often more than one factor contributes to the
  variation in the dependent variable
• You control for other factors
• Definition of controlling the process of
  holding constant the influence of a third
  variable on the relationship between two other
  variables

4
Purpose of multiple regression

• To provide an estimate of the independent effect
  of a change in each IV on the DV
• To provide an empirical basis for predicting
  values of the DV from knowledge of the joint
  values of the IVs

5
Specifying the model

• Translating verbal theory into an equation
• For multiple regression the formula is
• Y ao b1X1 b2X2 bnXn e
• See assumptions of regression pg. 295

6
Problems with Multiple Regression

• Non-interval data
• Convert into a dichotomy (coding is 0 not
  having the value, 1 having the value)
• Use a system of dummy variables to do the analysis

7
Regression with Dummy Variables

• Select Index Category
• Category you are comparing against 0
• 2 Values 0 or 1
• 1 has characteristic
• 0 does not have characteristic
• Female (1) Male (0)
• Married (1) Unmarried (0)
• Use Recode to Create Dummy Variables
• Number of Dummy Variables (categories 1)

8
Regression with Dummy Variables
9
Go to SPSS

• Open NES2000.sav
• Recode gender, partyID

10
Multiple Regression Interaction Effects

• Linear and additive technique
• Assumes a linear relationship between IV DV
• Assumes that effect of one IV on DV is same for
  all values of the other IVs in model
• OK for additive relationships
• Problematic if effect of one IV depends on the
  value of another IV, interaction
• How would we test for such relationships between
  independent variables?

11
Multiple Regression Interaction EffectsExample

• Polarization Perspective
• Political disagreements are often more intense
  among people who are more interested in and
  knowledgeable about public affairs than among
  people who are disengaged or lack political
  knowledge.

12
Multiple Regression Interaction EffectsExample

• Presume
• People who are Pro-choice would view feminist
  movement more favorably than non pro-choice
• But this difference will be greater among people
  with higher level of political knowledge
• Thus, strength of the relationship between
  abortion opinions and evaluations of feminists
  will depend on the level of political knowledge

13
Multiple Regression Interaction EffectsExample

• We want to specify a regression equation that
  does 3 things
• Estimate the mean difference on feminist between
  people with pro-choice and non-pro-choice
  opinions
• Estimate effect of political knowledge on DV
• Adjust additive estimate, based on value of
  political knowledge

14
Multiple Regression Interaction EffectsExample

• We want to specify a regression equation that
  does 3 things
• Estimate the mean difference on feminist between
  people with pro-choice and non-pro-choice
  opinions
• Dummy Variables
• Estimate effect of political knowledge on DV
• Multiple Regression
• Adjust additive estimate, based on value of
  political knowledge
• Interaction Term

15
Multiple Regression Interaction EffectsExample

• Interaction Term
• Multiply one IV by other IV
• permit polknow
• 0 on Permit (non-pro-choice)
• 0 for interaction term
• 1 on Permit (pro-choice)
• magnitude of interaction variable will increase
  with political knowledge
• Use Compute to create interaction variable
• Y a b1permit b2polknow
  b3(permitpolknow)

16
Deciding what type of analysis to do