Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
31 Cards in this Set
- Front
- Back
|
A Quasi experiment |
is like an experiment because it tries to establish a causal relation between one variable and another. But, it is not an experiment because it does not directly manipulate a variable. It tries to isolate a causal influence by selection rather than manipulation. Rather than causing X to vary from one condition to another, we select cases in which X does vary. There is no random assignment, there can be many confounds because of less control.In order to say something about causality, you have to look out for threats to internal validity (confounds). Example: The Bridge Experiment https://www.youtube.com/watch?v=Ji_osZc7z5Q |
|
|
Non-equivalent control group |
This is a control group identical to the other group except there is no treatment and no random assignment. The subjects were already part of that group naturally. https://www.youtube.com/watch?v=3lg_S1_nghg 3:27 |
|
|
Time-series design |
rather than comparing different groups, it looks at the same subjects but at different times (within-subjects). |
|
|
Wait-list control groups: |
everyone receives treatment, no random assignment. |
|
|
Threats to Internal Validity: |
• History • Maturation • Testing • Subject loss • Selection (systematic differences betweenconditions, not random) • Regression • Many other confounds |
|
|
Regression, |
It's statistical artefact that can occur when you are selecting based on Extreme scores. It's caused by random measurement error and/or natural random error (over time). A specific case is "Regression to the mean". Example: mood fluctuates over time, and at the lowest point someone will look for psychological help, which will turn the mood back to the mean(regression). |
|
|
Single-subject and Small-N Experiments |
Different types: • Time-series design • ABAB or reversal design • Multiple baseline design Main differences with case studies: • more control • more systematic • more often quantitative • more often hypothesis testing Characteristics of Small-N designs: • each subject is treated as a separate experiment, within-subjects design • reliability is assessed by replication • individual's subject's data are presented,variability is shown by comparison between subjects. Small-N designs must be careful of carryover effects such as fatigue and practice. |
|
|
ABAB / Reversal Designs |
This type of design depends on repeated measurements of behavior under at least two conditions (control +intervention). If there is a change after the intervention begins, and if nothing else in the situation has changed, it is likely that the intervention was responsible for the change. To test that conclusion, we can reverse the design. Example: A: Baseline B: Treatment A: no treatment B: treatment |
|
|
Intervention Study |
is the treatment effective for each individual? Positive effect on group level doesn't say anything about effect on a specific individual. To determine the effectiveness of a treatment for a specific individual, you have to take into account the amount of natural variability of complaints in this individual. |
|
|
a quasi experiment |
• Interested in causal relation indep. -dep. Variable • No random assignment (possible) • Less control (possible) • Many confounds • A good design increases internal validity Often selection of existing groups is used |
|
|
What do we need for causality? |
1. Is there an effect in the first place? (is there a significant difference?) 2. Did the cause take place before the consequence? (since we can’t travel in time) 3. Plausible alternative explanations (what about confounds?) |
|
|
longitudinal designs. |
Experimental group M1 -> X -> M2 2 measurements over time. Long-term designs with multiple measurements are known as longitudinal designs |
|
|
Disadvantages of longitudinal approach! |
Plausible alternative explanations? • Did the children learn the alphabet somewhere else? • Did the children who didn’t perform very well drop out? • Didn’t the children develop by themselves? • Aren’t the children more comfortable/confident at the 2nd measurement? • Etc. |
|
|
Non-equivalent control group with pre- and post test |
• Smart design! • Standard design in quasi-experimental research • This way you can control for many confounds |
|
|
threats of internal validity |
• History - Didn’t the children learn the alphabet somewhere else? • Maturation - Didn’t the children develop by themselves? • Testing -Aren’t the children more comfortable at the second measurement? • Subject loss Did the children who performed less well drop out? • Selection What about a priori differences between the groups? • Regression • And of course other confounds! |
|
|
history |
External influences on the dependent variables that coincide with treatment E.g., children go to school halfway through the experiment! |
|
|
Maturation: spontaneous development |
Changes in the dependent variable(s) that occur “spontaneously”, mostly by growing up and aging. Spontaneous remission in e.g. depression is also a form of maturation! |
|
|
Testng: re-test effects |
Familiarity of the subjects with the test increases the score. E.g., 2nd time you take a test you score higher because you already know the questions |
|
|
Subject loss |
Specific subject loss during research E.g., drop-out during longitudinal research about the effects of psychotherapy: the most depressed patients don’t show up anymore (loss of interest?). |
|
|
Selecton |
Systematic (not random!) differences between conditions that are not due to the treatment. E.g., gender, ethnicity, intelligence, personality, age, etc. |
|
|
multiple baseline design |
A multiple baseline design is a style of research involving the careful measurement of multiple persons, traits or settings both before and after a treatment. see this well |
|
|
reversal design |
ABAB design |
|
|
Selection vs. assignment of subjects |
-- |
|
|
limitations of quasi experiments |
-- |
|
|
absolute threshfold |
the faintest tone that we can just barely hear the lightest touch we just barely feel ... below some critical value, an input or stimulus is just too faint to be percieved at all |
|
|
threshold estimate |
the point at which detection occurs 50% of the time is taken as the threshold estimate |
|
|
how do we establish reliability and generality in small-N research? |
by replicating |
|
|
stable baseline |
|
|
|
multiple baseline |
|
|
|
reversal deign |
|
|
|
Cronbach’s Alpha |
The reliability of the scale https://www.youtube.com/watch?v=PCztXEfNJLM 1:53 |
