Transcript

III. Research Design Part I: Experimental Designs

Hypotheses A hypothesis is a testable statement of causal

relationship between two variables, derived from theory

Must specify a relationship between an independent and dependent variable

Clear, specific, amenable to empirical testing, value-free (F-N & N)

Research Design “The program that guides the investigator in

the process of collecting, analyzing, and interpreting observations. It is a logical model of proof that allows the researcher to draw inferences concerning causal relations among the variables under investigation” (F-N and N).

Relationships (Covariation) between Variables

Relationships between variables: Two variables are related to one another (i.e. are correlated) if one or more values of one variable tend to be associated with one or more values of the other variable.

Directional Relationships Apply to cases where the values of the IV and

DV are orderable (directional) variables Positive relationship:

As one’s education level increases, the frequency of voting increases

There is a positive relationship between one’s education level and voting frequency

Directional Hypotheses (cont’d) Negative relationship:

As the number of hours of negative ads watched increases, the frequency of voting decreases

There is a negative relationship between exposure to negative advertising and one’s voting frequency

Alachua

BakerBay

Bradford

Brevard

Broward

Calhoun

Charlotte

Citrus

ClayCollier

Columbia

Dade

DeSoto

Dixie

Duval

EscambiaFlagler

Franklin

Gadsden

GilchristGulf

Hamilton

Hardee

Hendry

Hernando

Highland

Hillsborough

Holmes

Indian RiverJackson

Jefferson

Lafayette

Lake

Lee

LeonLevy

Liberty

Madison

Manatee

MarionMartin

Monroe

NassauOkaloosa

Okeechobee

OrangeOsceolaPalm Beach

Pasco

Pinellas

Polk

Putnam Santa RosaSarasota

Seminole

St. Johns

St. Lucie

Sumter

Suwannee

Taylor

Union

Volusia

Wakulla

Walton

Washington

.05

.1.1

5.2

Ave

rage

Mon

thly

San

ctio

n R

ate

-3 -2 -1 0 1 2Local Conservatism (Initiative-Based)

Average Monthly Sanction Rate Fitted values

Non-Directional Hypotheses Appropriate for variables that are not orderable Hypothesis describes comparison among categories Examples

Men have greater levels of support for President Bush than do women

Whites are most likely to be Republican, while African-Americans are most likely to be Democrat

Research Design and Causality Relationships between variables: Two

variables are related to one another (i.e. are correlated) if one or more values of one variable tend to be associated with one or more values of the other variable.

Causal relationship: A relationship in which one variable directly causes/explains the other variable.

Establishing Causality (F-N&N) 3 Criteria (Evidence Needed) for Establishing

Causality Covariation (X is correlated with Y) Time Order (X precedes Y in time) Nonspuriousness (The observed relationship

between X and Y is not spurious)

Spurious Relationship A relationship between two variables that is

presumed to be causal, when in fact it is not

An observed relationship between X and Y is said to be spurious (or partly spurious) if there exists a third variable Z (a “control variable”) which is both a cause of Y AND is correlated with X.

Spurious Relationship

YX

Z

(Presumed Causal Relationship)

(True Causal Relationship)

Example of Spuriousness

Gender and Corruption – Causal or Spurious?

http://www.iq.harvard.edu/blog/pb/2005/10/sex_and_corruption.html

The Democratic Peace: Causal or Spurious?

WarDemocracy

Z(???)

Experimental Designs1. Select a sample2. Randomly assign subjects into 2 or more groups. The number of groups is equivalent to the number of values of the independent variable(s).3. Observe (measure) DV for all groups (if design includes pretest)4. Introduce the stimulus (IV)5. Observe (measure) DV for each group6. If the change in the value of the dependent variable varies significantly across groups, then we conclude that X Y

Key distinguishing features of an experimental design: Randomization and Manipulation of IV by the researcher (when introduced and to whom)

Experimental Designs

A bunch of people

TreatmentGroup

ControlGroup

RandomAssignment

RandomAssignment

Measurethe DV

MeasureThe DV

(PRE-TEST)

“Stimulus”

“Placebo”

Introducethe IV

TreatmentGroup

ControlGroup

Measurethe DV

(POST-TEST)

Measurethe DV

Simple Experimental Designs2-Group Pretest - Posttest Design (Classical or

“Simple” Experiment) R O1 X O2 R O3 O4

Simple Experimental Designs2-Group Pretest - Posttest Design (Classical or

“Simple” Experiment) R O1 X O2 R O3 O4 OR R O1 X1 O2 R O3 X2 O4

Experiments and Causality Correlation?

Experiments and Causality Covariation? Comparison of two or more groups (on

dependent variable) experiencing different levels of exposure to the causal (explanatory) variable (X). This establishes covariation.

Experiments and Causality Time Order?

Experiments and Causality Time Order? The introduction of the independent variable

(“stimulus”) is manipulated by the researcher to insure that changes in IV precede changes in DV.

Experiments and Causality Spuriousness?

Experiments and Causality Spuriousness? Random assignment insures that rival

hypotheses are ruled out, thus eliminating the threat of spuriousness. (How?)

Experiments and Causality Spuriousness? Random assignment insures that rival

hypotheses are ruled out, thus eliminating the threat of spuriousness. (How?)

Use of “matching” as a strategy to control for rival explanations

Simple Experimental Designs 2-Group Posttest Only Design R X O1 R O2

Simple Experimental Designs 2-Group Posttest Only Design R X O1 R O2 OR R X1 O1 R X2 O2

Other Types of Experimental Designs Multiple Group Pretest - Posttest Design

R O1 X1 O2R O3 X2 O4R O5 X3 O6R O7 X4 O8

Multiple Group Posttest Only DesignR X1 O1R X2 O2R X3 O3R X4 O4

Other Types of Experimental Designs Solomon 4-Group Design

R O1 X O2

R O3 O4

R X O5

R O6

Extensions of the Classical Experiment Multiple observations over time

R O1 X1 O2 O3……

R O4 X2 O5 O6……

Extensions of the Classical Experiment Factorial designs - each group represents a unique combination of

values on two (or more) different variables.

For two independent variables X and Z (where X and Z each take on two possible values):R O1 X1, Z1 O2R O3 X2, Z1 O4R O5 X1, Z2 O6R O7 X2, Z2 O8

Factorial designs allow the researcher to test for an interaction effect: Two independent variables interact to affect a dependent variable if the effect of one variable depends on the value of the second variable.

Zilber and Niven (SSQ) Table 1: 2-Group Posttest Only

R X1(black) O1

R X2(A-A) O2

Zilber and Niven (SSQ) Table 3: 2X2 Factorial Design

R (black/liberal) O1 R (A-A/liberal) O2 R (black/conserv) O3 R (A-A/conserv) O4

To see how the effect of racial label varies as a function of ideology, we compare

O1-O2 to O3-O4

Zilber and Niven (SSQ) Table 3: 2X2 Factorial Design

R (black/liberal) O1 R (A-A/liberal) O2 R (black/conserv) O3 R (A-A/conserv) O4

Conclusion: The choice of racial labels does affect white attitudes toward blacks, but only among liberals.

1.What effect do A & B

have on O?

2. Is there an interaction

effect? (Explain)

Evaluating Research Designs:Internal Validity Internal Validity - the degree to which we can be sure that the

independent variable caused the dependent variable within the current sample

“Extrinsic Factors”: Selection effects with respect to recruitment/assignment of subjects (units) to treatment and control groups

“Intrinsic Factors”: Factors threatening validity that… occur outside the “laboratory” during the period of the study result from changes in, reactions to (or general ineffectiveness of) the

measuring instrument, or involve some type of reactive effect of observation

Evaluating Research Designs:Internal Validity Extrinsic factors:

Selection Important intrinsic factors include:

History Maturation Experimental mortality Instrumentation Testing Regression artifact Interactions with selection - e.g. “selection history” and

“selection maturation”

For each of the following intrinsic threats to internal validity, explain: What the specific threat means Whether or not (and why or why not) an

experimental design is protected from this threat

History? R O1 X1 O2 R O3 X2 O4

Maturation? R O1 X1 O2 R O3 X2 O4

Experimental Mortality? R O1 X1 O2 R O3 X2 O4

Instrumentation? R O1 X1 O2 R O3 X2 O4

Testing? R O1 X1 O2 R O3 X2 O4

Regression Artifact? R O1 X1 O2 R O3 X2 O4

“Selection History” and “Selection Maturation”

R O1 X1 O2 R O3 X2 O4

Evaluating Research Designs:External Validity External Validity - the degree to which the

results of the analysis can be generalized beyond the current sample/study. Can be maximized by: Using subjects (units) that are representative of

the population to which one’s theory applies Using a “laboratory” that is as close to “real life”

conditions as possible Field experiments

Applications

1. Iyengar, Shanto, Mark D. Peters, and Donald R. Kinder. 1982. “Experimental Demonstrations of the ‘Not-So-Minimal’ Consequences of Television News Programs.” American Political Science Review 76: 848-58.

2. Schram, Sanford F., Joe Soss, Richard C. Fording and Linda Houser. 2009.  “Deciding to Discipline: A Multi-Method Study of Race, Choice, and Punishment at the Frontlines of Welfare Reform .”  American Sociological Review, 74(3), 398-422.

3. Gerber, Alan S., Donald P. Green, and Christopher W. Larimer. 2008. “Social pressure and voter turnout: evidence from a large-scale field experiment.” American Political Science Review 102:33–48.

  

Assignment #5

(Due October 5): In approximately 2 single-spaced pages, answer the following questions.

1. For each of the three application readings, identify and diagram (using the notation in Frankfort-Nachmias and Nachmias) the specific type of experimental design employed by the authors.

2. Choose one of the three studies to focus on for this question. Evaluate the internal validity and external validity of the study you have chosen.

 

top related