Closed system
|
|
Important factors that influence environment is controlled by the experimenter
Confident that IV causes change in DV
Rule out rival hypothesis
Not always possible
High internal validity
|
| |
Open System
|
|
Participants can be influenced by a number of factors in which the researchers have little to no control
used when strong experimental designs are not possible
Use common sense and reasoning to interpret data
|
| |
Quasi-experimental designs
|
|
used to evaluate the impact of a variable on an ongoing process
|
| |
Correlational designs
|
|
describe the relationship between two variables
|
| |
Naturalistic Observations
|
|
describe an ongoing process in its natural setting
|
| |
Quasi experimental designs:
Time series design
|
|
Within-subjects design
Useful when interested in the effects of an event that has happened to all of the population being studied
single-group pretest-posttest design, weak design
|
| |
Quasi experimental designs:
Interrupted time series design
|
|
several pretest and posttest measurements
Can add reversal of phenomenon to add confidence to results
|
| |
Quasi experimental designs:
Longitudinal designs
|
|
Time is one of the independent variables (related to time series design)
same participants assessed over multiple time points but nothing is introduced between any of the time points
Potential confounds
|
| |
Quasi experimental designs:
Nonequivalent control group posttest- only
|
|
Exp. group X 01
Non equivalent control group O1
Used to compare effectiveness of some program when random assignment is not possible
low internal validity - bc groups arent similar and don`t know how they differ because no pretest
Weak design
|
| |
Quasi-experimental Designs:
Nonequivalent before after design
|
|
Pretest provides baseline info
Very informative when random assignment is not possible
Used in educational research
Compare the differences between before and after scores
|
| |
Test writing productivity, u of s tests program, u of r control group. Whats happening Threats to internal validity:
|
|
Both are increasing, so something else is affecting it
Maturation- could just be because one program is writing more than the other
History`- because of pretest people want to get better, so they use feedback to write better
|
| |
|
|
No increase with u of r but there is an increase with u of s
Might conclude that the program is effective because of the increase
Could be because of ceiling effects - u of r might already be at the top
Regression effects towards the means
|
| |
|
|
No increase in u of r but in crease in u of s started out similar
More confidence in the effectiveness of the program
Selection by maturation: no random assignment, so u of s might attract students that are stronger writers
|
| |
|
|
Strongest evidence that program is working
Most confidence in program
|
| |
Quasi experimental designs:
Optimizing nonequivalent before after designs
|
|
try to get control groups as similar as possible to experimental group
collect as much demographic info as possible
knowing how groups differ increases confidence in results
|
| |
Quasi-experimental designs:
Retrospective and ex post facto designs
|
|
Attempts to use empirical procedures to study something that happened in the past, try to suggest meaningful relationships
Weak forms of inference but can be used in testing alternative hypotheses
More in correlational studies
Always threats to internal validity
|
| |
Quasi-experimental designs:
Comparative time series design
Exx.
|
|
asses two or more variables at different time periods to see how one variable relates to changes in the other
Compare fatalities per 1000 canadians to the number of provinces banning cell phone use while driving
|
| |
Why do we use quasi-experimental designs
|
|
to study interesting events in real world settings that can not be studied in a lab
enhances external validity
ethics
|
| |
Head start study
-Early childhood prgram in the US for preschoolers from low SES homes
-Comparison group identified using achievement scores
matched group not similar SES backgrounds
What kind of design
|
|
Non-equivalent pretest posttest design
|
| |
Employees at an electronics company assessed for job burnout before and after a scheduled vacation
level of burnout was assessed before, during and after the 2 week vacation
What kind of design
|
|
Interrupted time-series design
|
| |
|
|
not maturation because no stead inc or dec
maybe selection by history effects- vacation might be in summer so people happier in summer
Attrition- might be people who are less burned out doing the test on vacation
|
| |
Correlational studies
|
|
Describes a relationship between two variables
does not imply causality, doesnt attempt to show how one influences the other
|
| |
Correlation coefficient
factors that influence it
|
|
stat used for establishing relationship degree
factors:
restricted range, outliers, reliability of measures
Lack of correlation rules out possibility of causality
|
| |
Advanced Correlational strategies
|
|
Regressional analysis
Cross-lagged panel and structural equation analysis
Multielevel modeling
factor analysis
|
| |
Naturalistic Observations
|
|
Pay attention and observe
Useful in complex situations
useful when little is known
|
| |
Problems when making observations
|
|
Reactive behaviour- participant is influenced by presence
Unobtrusive observations- undectected
Selective perception- when observers record things they expect to see
Fatigue and boredom
|
| |
When anaylzing observations what do you do
|
|
review data several times
look for patterns
evaluate atypical behaviour
evaluate any theoretical ideas you may have had
|
| |
Factors to be considered when decideing whether you or not you should conceal your identity
|
|
observing highly reactive behaviors
if behaviour will be different then use some consealage
If there is much to be gained by personally experiencing and wont be significantly influenced then participate
|
| |
Naturalistic observation advantages and disadvantages
|
|
Adv.:
Describe behavious as it naturally occurs
sutdies behavioural processes over a period of time
DIsadv:
tend to be qualitative
representativeness of sample may be compromised
doesn`t provide info about how one variable influences the other
|
| |
Four categories of Behavioural research
|
|
Descriptive
correlational
quasi-experimental
experimental
|
| |
