This preview shows page 1. Sign up to view the full content.
Unformatted text preview: Quasi and Hybrid Experimental Designs I NonEquivalent Group Design
(NEGD) What distinguishes an experiment from a quasiexperiment? Nonrandom assignment to experimental and control groups (in book) Direct manipulation of the experimental effect (not in book) Illustrative example:
Relationship between physical activity on breast cancer Background:
Want to design a physical activity intervention to see if it reduces reoccurance of breast cancer You are a physician and you have access to two support groups You want to set up a study to test the effectiveness of your intervention What will your study look like? What is your dependent variable? What is your independent variable? How would you set up the design? What kind of sampling are you using? Nonequivalent group design Meant to deal with the issue of Nonrandom assignment N O X O N O O Deals with the reality of two separate and independent pools of individuals in the "Control" and "Program" groups What Does Nonequivalent Mean? Assignment is nonrandom. Researcher didn't control assignment. Groups may be different, may come from different populations! Group differences may affect outcomes. The Design... Physical Activity and Cancer example... 2 Support Groups 1 Given Intervention (Experimental/Program) 1 Not Given Intervention (Control) Weakness in design is based on a priori differences in groups at start Differences may have a significant impact on the overall outcome 80 75 70 65 60 55 50 45 40 35 30 Graph of Means
Comparison Program Pretest Posttest Comp Prog pretest MEAN 49 54 posttest MEAN 50 65 pretest STD DEV 6.9 7.0 posttest STD DEV 7.5 7.3 Summary NEGD is one of the most common quasi experimental designs Strength of result is dependent upon nature of result Note: Distinction between experimental and quasiexperimental design is not always clear Mostly due to random assignment and selection Summary Design Goals Observe in a systematic way Draw conclusions about relationships among So far: variables Experiment, quasiexperiment, naturalistic Establishment of causality? Distinction between an experiment and a quasi experiment?
NEGD No control over assignment of experimental and control group Review of Designs Thus Far... When do you not have control over assignment of condition? Most things in behavioral research: Clinical studies: Alzheimer's vs. control adults ADHD vs. control children Leukemia survivors vs. control children Demographic variables Ethnicity Gender Socioeconomic level Job status To this point we have examined studies where: Dependent variable is continuous Independent variable is dichotomous or polychotomous When is this not possible or desirable? What do we do then? Correlational Design Cases where independent variable and dependent variable are continuous? Relationship between social support and recovery from Leukemia Correlational design Continuous independent and dependent variable R=.5 Correlation Ranges from 1 to 1 O means no relationship 1 means negative relationship 1 means positive relationship Smoke 5 packs a day, have poor lung capacity Exercise 3xper week, have good lung capacity ProxyPretest Most appropriate when the intervention starts after the pretest!!! Example: Interested in interventions that maintain quality of life in demented elderly ALMOST NEVER HAVE PREDEMENTIA info Design
N O1 X O2 N O1 O2 Looks like nonequivalent groups design but one big difference! "preintervention" assessment is retrospective Two ways it can be retrospective #1: Ask people to "think back" to what happened to them earlier Disadvantages? Why would you want to even bother? Two ways it can be retrospective #2 use archival data that precedes intervention E.g. medical records, school records Advantages? Disadvantages? Bottom Line NEVER CHOOSE TO DO THIS!!! But sometimes stuck with it because of Poor planning Nature of the sample Nature of the research question Summary Experimental design: Have random assignment to groups and control over condition Quasiexperimental: Have nonrandom assignment and potentially no control over condition Correlation: Have quantitative independent and dependent variables Distinction between these categories is not always clear ...
View Full Document
- Spring '07