Quant: Lecture 3 Flashcards
Describe a field experiment
It’s an experiment that occurs in a natural setting and the experimenter still manipulates the IV. However, it’s hard to control extraneous variables.
Describe quasi experiments
It occurs in a natural setting and the IV is naturally occurring. However, you can’t control extraneous variables that could affect the DV. There are three types of quasi experiment: pre-test post-test single group design, one group post test with no baseline or control (beforehand study), post test with no equivalent group, no baseline and a control that isn’t randomly assigned and a pre-test post-test no equivalent with baseline and control with no random assignment.
Describe a laboratory experiment
The IV is manipulated and there is a lot of control over extraneous variables.
How do you design an experiment?
Firstly, you start with observation. Then you formulate a null and an experimental hypothesis which is either two-tailed (no direction) or one-tailed (direction stated). After this, you measure the hypothesis to either falsify or verify it. To do this, you pick a design which is either a correlation or experiment. Next, you formulate your variables and pick either a unrelated (between subjects) or related (within subjects) design.
List 2 advantages and 1 disadvantage of an unrelated sample
Advantages: There is less room for learning and you can make group comparisons
Disadvantages: If participants aren’t randomly assigned then it could be biased.
List 1 advantage and 1 disadvantage of a related sample
Advantage: There is less room for assignment bias
Disadvantage: There is more room for learning.
How many trials should you do before an experiment?
How should carry out your trials?
You should try and do 6 trials, however, if there are less individuals then you should do more trials per individual. There are two ways of doing trials; blocked trials which involves the same participant in the same condition but the number of trials in each condition is counterbalanced, there are possible order effects and mixed trials which involves semi-randomised trials meaning the participants are randomly assigned to a condition, there is less chance for order effects and the amount of trials are counterbalanced, however this can be too complicated meaning the effects can vanish.
When should you use parametric statistics?
Give an advantage and disadvantage
When there are more than 15 individuals and the data is normally distributed.
Advantage: It is more powerful
Disadvantage: There are lots of criteria that needs to be fulfilled.
When should you use a non-parametric statistic?
Give an advantage and disadvantage
When there are less than 15 individuals and the results aren’t evenly distributed.
Advantage: There isn’t a lot of criteria that needs to be fulfilled
Disadvantage: It’s less powerful.
When should you use the non-parametric chi squared?
When your data is nominal, unrelated and experimental.
When should you use the non-parametric spearman’s rank?
When your data is correlational and numerical.
When should you use the non-parametric Wilcoxon?
When your data is related and numerical.
List 4 disadvantages of the experimental method
Demand characteristics, experimenter effects (you may influence the participant), ecological validity and mundane realism.
What is mundane realism?
When an experimental experience is similar to an everyday one, however, there can still be high experimental realism even if there is low mundane realism.
Describe cohen’s d and how you calculate it
Cohen’s d measures the effect size of your results by comparing the means and SDs. It measures the strength of the difference between conditions. A small effect; 0.2, medium; 0.5, large; 0.8. You measure it by doing: the mean of condition 1 minus the mean of condition 2. Then you add the squared SD for condition 1 with the squared SD for condition 2. Next, you divide this result by two and then square root the answer. Finally, divide the minus means by this answer. This will tell you the number of SDs by which the means differ. The more the means differ and the smaller the SDs, the larger the effect size. A large effect size doesn’t make the results meaningful.
