Ray Block, Jr.
PS 585
Research Methods

Today’s Blueprint
Last Class

• Experimental Research Designs
• The logic of experiments
• Types of experiments
• Strengths and weaknesses

• Non-Experimental Research Designs
• Experimental vs. non-experimental designs
• Types of non-experimental designs

Telling Them Apart

• Remember from last class…
• Active ingredient in experiments:
1. Control = Researcher decides what the treatment is and how it is carried out
2. Random Assignment = Researcher decides who gets which treatment and how

• Experimental Designs:
• Test relationships between variables by controlling or manipulating subjects and conditions

• Non-experimental designs:
• Test relationships between variables without  controlling or manipulating subjects and conditions

Some examples:

• [Non-experimental] Time Series
• Cross-Sectional
• Case Study
• Panel

1) Non-Experimental Time Series Designs

• What are they?  Multiple measures of the dependent variable are taken before and after the introduction of independent variable
• What purpose do they serve? Uses time as a natural experiment:
• DV at time1 (Quasi control)
• Event occurs
• DV at time2

• Why do we do them? When you need to control for the effects of time
• When should we use them? When you believe that a particular event influences your dependent variable
• What types are there?
• Extended time-series
• Extends the range of time measured before and after independent variable
• Controls for maturation effects
• Controlled time-series
• Combines time-series data with a comparison group
• Controls for history, maturation, and test-retest effects

2) Cross Sectional Designs

• What are they? Measurements of the independent and dependent variable are taken at the same time
• What purpose do they serve? Uses naturally occurring differences in the independent variable to create “quasi” experimental control groups
• Why do we do them?
• Representative
• Realistic
• Cheaper

• When should we use them? When we want to examine the relationship between variables (correlation, not causation)
• What types are there? (Think surveys)
• Phone: Cheap and fast. But verbal instructions and response alternatives may be hard to remember. + Who is home?
• Face-to-face: Expensive but maximum control
• Mail: Cheap, but response bias issues
• Internet: Cheap but response bias also; less control

3) Case Study Design

• What are they? Intense study of a single case/scenario or a several cases/scenarios
• What purpose do they serve?
• Exploratory
• Descriptive
• Explanatory

• Why do we use them? Investigate a contemporary phenomena
• Boundaries between phenomena and context are not clearly defined
• Multiple sources of evidence

• When should we use them?
• Confirm the [causal] relationship between variables
• When quasi-experiments exists

• What types are there?
• Single Case Study
• Example: Pointing out deviant cases
• Multiple Case Study
• Example: Making comparisons between:
• Different cases over time
• Similar cases at different times/situations

4) Panell Study Designs

• What are they? Measures of variables are taken on the same units of analysis at multiple points in time
• What purpose do they serve? Examine before and after conditions by examining the same sample over several periods
• Why do we do them?
• To establish over-time trends
• To explain changes in trends

• When should we use them?
• When data is consistently collected over time on the same sample or unit of analysis
• Example: census data

To Recap: See Table 5-5 (p. 147) in Johnson et al. text