Is there a relationship between factionalization of a country and the time in between elections?
Factiona lizatio n
Constant
.^/
laining length of coalition governments
std.
95% confidence
Coef.
Error
interval
-1.35
-0.381
-2.31
o.49
0.166
-o.39
325
R2 0.

POLS 3800 Introduction to Methods of Political Research
Dr. Judd Thornton
Office: 1013 Langdale Hall
Office Hours: M-F 9:3010:30
Email: jrthornton@gsu.edu
Goal of course
The goal is to introduce students to the methods used by political scientists to try

Study Guide
POLS 3800
Regression
Basics
Why do we want a probabilistic model?
= $ + & +
Goals: hypothesize relationship; use sample to estimate model; assess usefulness of model
Sum of squared errors: +),& )*
Whats ?
Assessing Usefulness
1 Assessing

Multiple Regression: Interaction Terms
May 26
Interactive Relationships
Some relationships are conditional.
The relationship between X and Y depends on Z .
The impact the slope depends on the value of Z .
Quality = 0 + 1 Experience + 2 Female + 3 (Exp.

Multiple Regression: Dummy Variables
May 26
Categorical Independent Variables
What if were interested in the impact of gender?
Or region?
Or race or ethnicity?
Income = + 1 Gender
Views Toward Hillary Clinton
Does gender matter?
Hillary = 0 + 1 Income

Bivariate Regression
May 24
Regression
Move to model building.
Still testing a hypothesis.
Regression
y = mx + b
Linear, deterministic model.
Deterministic: y can always be found if we know x.
Social world is not deterministic.
y = + x + Random Error

Multiple Regression
May 25
Linear model
We believe that education affects political participation. We
model:
Participationi = 0 + 1 Educationi + ui
1 is the impact of education on participation.
We cannot establish a causal relationship.
We can rule ou

Multiple Regression
May 25
More on Omitted Variable Bias
If we exclude z,
1 = 1 + shared variance with z
We over or underestimate the impact of x on y .
We call this bias.
Hypothetical Example 1
Imagine the following is the true process:
Score on testi =

Bivariate Regression
May 25
Linear model
A reminder of what were doing:
We hypothesize a model:
yi
= + 1 xi + ui
where 1 is the impact of x on y.
Were interested in explaining the part of the world thats
systematic.
We want to find generalizable feature

Recap
May 27
Dummy Variables
Why use a dummy variable?
How do we code dummy variables?
Zero and one.
Dummy Variables
The impact of Maymester, seven-week summer courses, or
Fall/Spring on retention?
Two dummy variables.
k 1 dummy variables (where k i

The Big Picture
May 27
Lets Avoid being Casual about Causality
Goal is to explain systematic portion of the social world.
Avoid (exclusively) relying on anecdotes.
Does presentation of news influence attitudes?
We could study this experimentally!
How