Unit 8 Flashcards
Research procedure which does not meet the requirements of a true experiment
Quasi-experiment. Because of this reduced level of control, inferring causality in a quasiexperimental design cannot be achieved with the same level of certainty as a true experiment
The independent variable’s of an experiment are often called
Factors.
For example, an experiment on the effect of human handling on the socialization of tiger cubs, tiger cubs would be either handled are not handled by the people as they were reared. Handling was the independent variable, the factor of this experiment.
In an experiment a particular value of an independent variable
Level.
An independent variable always has at least two levels. In the experiment of the tiger cubs, the two levels would be handling versus no handling.
A group or treatment in an experiment
Condition
In the tiger handling experiment, experimental conditions are the same as the groups. Anyone tiger in this example was either handler not handled, each tiger experienced only one treatment or condition
Another word for a condition of an experiment
Treatment
To particular elements of design provide control over so many different threats to validity that they are basic too good experimental designs:
The existence of a control group or a control condition and the random allocation of subjects to groups
Changes in a subject’s performance resulting from the position in which a condition appears in an experiment in a within subjects design
Order effects
And example of an order affect is the warm-up or practice effect that often occurs in experiments on learning. Whichever condition is presented first will show poor performance than later conditions simply because the subject has not warmed up to the task
Changes in the subjects performance resulting from interactions among the conditions themselves in it within subjects design
Sequence effects
For example, in an experiment done judging the heaviness of lifted weights, there’s likely to be a contrast affect such that a lightweight will feel even later if it follows a heavy one, and vice versa
Controlling for order and sequence effects by arranging that subjects experience the various conditions in different orders
Counterbalancing
What is the basic strategy for achieving control with within subject designs?
Two basic strategies are available for controlling order and sequence effects. The preferable one is to arrange the order of conditions in such a way that order and sequence effects are controlled within subjects. When this is not possible, you must control for order and sequence effects between subjects.
Controlling for order in sequence affects within subjects as possible in each subject receive each condition. Randomization can be used if each condition is given several times to each subject or when a sufficient number of subjects will be tested such that one particular sequence is unlikely to have much influence on the outcome.
Control procedure in which the order of conditions randomized but with each condition being present it only once before any condition is repeated
Block randomization
Block randomization is particularly useful if you want to present each condition at least twice and your experiment requires two or more sessions. So it is most useful when conditions are presented several times to each subject
Method of controlling which conditions are presented in order the first time and then in reverse order
Reverse counterbalancing
Useful when relatively few subjects will be tested and you have several conditions that can be presented only a few times
ABBA ABCCBA
When is reversed counterbalancing ineffective in controlling for order effect
When a variable has a large effect in the early part of the experiment but a smaller effect later on. The best example is a warm-up, or practice effect that may occur in the early part of an experiment and be less important later
Reverse counterbalancing me do an incomplete job of controlling sequence effects in an ABCDDCBA experiment: the B condition follows A once and C once, but never follows D or itself
How do you control for order and sequence effects within groups
If present in each condition enough times to randomize the order is not possible, or if counterbalancing within subjects does not seem appropriate, you must leave order and sequence confounded with conditions within subjects. Then you must control for ordering sequence between or across subjects, essentially within the group.
You would use complete counterbalancing within the group because each condition occurs an equal number of times and each rank order and follows every other condition equal number of times. The disadvantage of this method of counterbalancing is that as the number of conditions increases, the number of orders required increases geometrically.
If you give up the requirement that each condition follow every other condition and equal number of times, you would be controlling for order but not for sequence of conditions. This type of incomplete counterbalancing is called the Latin square technique, after the agent positive finding ways to arrange a number of letters in the matrix such that each letter occurs once and only once in each row and column. Disadvantage is order is not controlled
You can control for sequence effects of the immediately preceding condition by using particular sets of Latin squares known as balance squares. In the balance Latin Square, each condition is immediately preceded once by every other condition
Control procedure in which each subject experiences each condition in a different order from other subjects
Latin square
A disadvantage of the Latin square technique is that sequence is not controlled, thus sequences always perfectly confounded with order in this particular Latin square
The advantage of the Latin square technique over complete counterbalancing is that a permits greater flexibility in choosing a number of subjects to be tested
