Chapter 11 - Confounding and Obscuring Variables Flashcards
For single group pretest-posttest designs, there are six possible thereat that can affect internal validity, what are they?
Maturation
History of effects
Regression to the mean
Attrition
Testing threats (test repetition effect)
Instrumental test
Maturation
Participants automatically become better at tasks over time
History of effects
Between pre- and post-test something happened to most participants, making it unclear whether the change was due to the manipulation or something else
It is wise to sue a comparison group for this as well
Regression to the mean
An experimental group changes over time because the most extreme cases are systematically removed during the study, so that they are not included in the follow up measurement
Why is attrition an issue when it happens systematically?
When only a certain type of participant drops out
Testing threats (test repetition effect)
Is a type of sequence effect where an experimental group changes over time, because related testing affects participants
After a number of tests, they may be tired or bored, so they will perform less well
When is it best to use a posttest?
Using a posttest only with a control group or to use different variants of the test otherwise
Instrumentation test
When a measurement instrument changes over time
Criteria have changed at posttest (like observers)
Different tests (such as observers) used you don’t know where the difference comes from
What is the difference between testing and instrumentation?
With testing, the participants are changed and with instrumentation the measuring instrument
What are the threats to internal validity for all types of experiments?
Observer bias
Placebo effect
Observer bias
Expectations colour the observation
Demand characteristics (also observer effect)
Participants try to guess what the researcher wants to know and act/respond accordingly
How to reduce the problem of demand characteristics?
Double blind study
Masked design
What is a masked design?
Researchers do not know which group the participants are in
Placebo effect
Participants in the experimental group improve because they believe in the effect of the therapy or medicine they received, even if ti is a placebo
What are the causes of a null effect/null result?
Not a large enough between groups difference due to weak manipulations, a not sensitive enough measurement instrument or ceiling and floor effects
Ceiling and floor effects
With a ceiling effect, all the scores clustered together on the high side and with a floor effect, that cluster is on the low side
What can the cause be of a ceiling and floor effect?
Can be caused by a badly manipulated independent or dependent variable
Within group variability (noise)
Can overshadow the effects
This can include measurement error, individuals differences, situational noise, and power
Individual differences
Participants differ in certain aspects such as motivation and ability
For example, reading a text
Situational noise
A noisy location can cause participants to become distracted
