Within-subject design
- A single group of subjects is exposed to all of the treatments, one treatment at a time
- Looking at the performance of each subject across treatments
Between-subject design
Each treatment is administered to a different group of subjects
The randomized two-group design
randomly assign your subjects to two groups and expose them to two different levels of independent variable, and take steps to keep the extraneous variables constant
→Compare the two means to determine whether they differ
Parametric design (The randomized multi-group design)
systematic variation of the amount of the independent variable (quantitatively e.g. different dose of medicine)
non-parametric design
The randomized multi-group design
manipulate your independent variable qualitatively (e.g. different kinds of medicine)
multiple control group design
The randomized multi-group design
when a single control group is not adequate to rule out alternative explanations
Matched group design
Matched pairs design
Determine a characteristic and test your participants for them now pair these who have the same results and assign them to a different group
Matched multi group design
requires finding matched subjects for each of your groups (so 4 groups = 4 matched subjects)
Mixed design
• A field was divided into several plots, and different plots received different levels of a given treatment. Each plot was then divided into subplots and each subplot received a different second treatment, thus each plot received all levels of fertilizer, but only one level of pesticide. Each plot are groups who all receive the same level of between-subjects variable. The subplots receive different levels of within-subjects variable to which all members of that group are exposed.
Advantages of within-subject designs
- Matching is on a high level so any differences across treatment cannot due to error variance arising from such differences (closest matching because you compare somebody to himself)
- More sensitive to the effect of your independent variable
- You can use fewer subjects
Disadvantages of within-subject design
- Require time to complete →might be difficult to find participants for this
- You could split up the experiment to different times but than you have the problem that some might leave the experiment because they wont show up to following treatments
- Carryover effect: When a treatment changes the behaviour of a subject these carry over all treatment and changes the whole performance of the subject (matched groups are a solution for this problem)
Advantages of between-subject design
+safes time and money
+relatively easy to analyse statistically
Disadvantages of between-subject design
- provides a limited amount of information about the effect of the independent variable (you can only compare the average)
- you don’t learn much about the nature of the relationship and function, in view of your independent and dependent variable
- sensitivity is limited to the effect of the independent variable (characteristics can influence the outcome so you can’t really determine the effect of the independent variable)
error variance
the variability among scores caused by other variable than your independent variable (e.g. age, gender, personality)
Interferential statistics
Show you the probability with which error variance alone would produce between group differences on your dependent measure at least as large as those you actually observed in your data. When its low it is statistically significant
Carryover effect
When a treatment changes the behaviour of a subject these carry over all treatment and changes the whole performance of the subject (matched groups are a solution for this problem)
Sources of carryover effect
- Learning how to perform a task because they already did one ore more
- Fatigue: when previous performance leads to fatigue it influences the performance in following treatments
- Habituation: When people get used to a treatment it can lead to reduced responsiveness
- Sensitization: Sometimes exposure to a stimulus ca increase responsiveness to another stimulus
- Contrast: when a subject can compare what they receive in one condition with what they receive in another. When it’s less it can decrease performance other way around (more money stronger performance/less money weaker performance)
- Adaption: If subjects become used to a certain situation e.g. low light in a theatre it can influence their following performance (in case of drugs it is called tolerance)
counterbalancing
- Assign the various treatment in different order for different subjects
- Your number has to be equal of different treatments k formula: k= kk-1k-2k-3 (e.g. three treatments 321= 6 so you need 6 participants) (when you need more subjects add always the same amount)
Latin square
You use as many treatment orders as you have treatments
Making treatment order an independent variable
- Different groups of subjects to different orderings of treatments
- Changing one treatment to a testing variable (factorial design)
- You can measure the seize of every carryover effect that may be present.
Taking steps to minimize carryover
- Psychophysical and human decision making experience are not affected by carryover
- Giving participants time to become familiar with the treatments before the treatments so adaption and habituation is done before the treatments
- Allow breaks between treatments
Within-subjects versus matched group designs
- If you have reason to believe that the relationship between subject variables an your dependent variable is weak, use a randomized-groups design
- Matched group design is better if you are concerned that carryover effects will be a problem
Types of within subjects:
• Single-factor two-level design:
→includes two levels of a single independent variable
→both receive both treatments but in switched order
• Single-factor multilevel design:
→a single group is exposed two tree or more levels of a single independent variable
