Você está na página 1de 5

1.

Compared to a between-subjects design, a within-subjects design has the benefit that it


is more likely to yield a large F ratio. The reason for this is that F is the ratio of betweensubjects variance (which is the effect of the iv on the dv) to within-subjects variance,
which is error. The main cause of error is variability among the subjects in each
condition of the experiment. Because a within-subjects design uses the same subjects in
each condition, there is less variability among subjects in the experiment. This means
that the error term, which is the denominator of the F ratio, gets smaller, and therefore the
F ratio itself gets bigger.
2. The Hawthorne effect is produced when receiving attention from the experimenter
causes subjects to behave differently than they would normally. For example, people
tend to do better at simple tasks when they are being watched, which is called social
facilitation, and they do worse at hard tasks because of the distraction factor. The only
way to control for the Hawthorne effect is to have a control group that is given attention
by the experimenter but that does not get the experimental treatment. That way, you can
look at the effect of that attention separately from the effect of the treatment.
3. Within-groups variance is the result of error.
4. When you graph the results of a 2 x 2 experiment, you can spot an interaction effect
by looking for non-parallel lines. These lines let you know that two groups of subjects,
corresponding to two levels of one of the iv, are responding differently to a change in the
level of the second iv.
5. A mixed design combines elements of a between-subjects and a within-subjects
design.
6. Ecological validity is the extent to which your research findings generalize to
conditions and groups of subjects other than the ones you studied. Demand
characteristics threaten ecological validity by causing the subjects to behave in ways that
they think the experimenters expect them to behave rather than how they would behave
normally. This creates an artificial situation that may not generalize to conditions other
than those found in the context of the research study.

7. If the F ratio is large enough, it lets us know that the results we obtained could not
have occurred by chance alone. We dont have a guarantee that our results are caused by
the treatment, as opposed to a confound, but at least we can rule out chance.
8. The Hawthorne effect is not related to experimenter bias because it has more to do
with the behavior of the subjects than of the experimenter. Even an experimenter who is
perfectly unbiased can create a Hawthorne effect simply by being in the same room as the
subject.
9. Counterbalancing is when individuals are assigned to conditions so that each condition
appears equally often and in different orders.
10. Experimenters test hypotheses in the following order: null, confound, research.
First, you see if you can rule out the null hypothesis. If you can, then you try to rule out
the effect of confounds so that you can lend credibility to your research hypothesis.
11. Demand characteristics are when subjects behave in the ways they think the
experimenter expects them to behave.
12. There are three null hypotheses in a 2 x 2 design, one for each of the two main
effects and one for the interaction of the two.
13. This question was poorly worded, for which I apologize. As a result, it is one of
three questions on the exam that I threw out. Basically, everyone got this one right. The
question asks what happens when subjects in different conditions are NOT treated
identically. The answer is that a possible confound is introduced. The only difference in
the way subjects in different conditions are treated should be the treatment itself. All
other variables should be held constant across these conditions.
14. A completely randomized design is a between-subjects design, by definition. You
cannot do random assignment of subjects in a within-subjects design because you have to
include all subjects in all conditions, which is NOT a random form of assignment.

15. A multileveled completely randomized design is analyzed using a single-factor


ANOVA. Even though there are multiple levels, you still have only one iv. Given that
you have more than two levels, you cannot use a t-test.
16. In a 2 x 2 design, an interaction effect occurs when the effect of one independent
variable depends on the level of the other independent variable. Just think back to the
example Ive used of the study of eating behaviors among obese and non-obese people.
The number of crackers people eat when theyre full depends on whether or not theyre
obese.
17. You do not want participants in your experiment to communicate with each other
because it produces a threat to internal validity called diffusion.
18. A true experiment must have random assignment, controls, at least two levels of the
iv, and a null hypothesis. It does not need to have a pretest.
19. You can have a significant interaction effect in a factorial experiment with or without
there being main effects for the two ivs.
20. When an iv has no significant effect, F = 1. That is because the between-subjects and
within-subjects variance are equal.
21. In a within-subjects design, subjects serve as their own control because every subject
takes part in every condition.
22. If you want to make sure that your results generalize to a specific population, you
want to be careful to choose subjects who are representative of the population to which
you wish you generalize.
23. A matched-subjects design is more sensitive than a between-subjects design, meaning
that it produces less within-subjects variance, or error. But it is not as sensitive as a
within-subjects design, which minimizes error by using the exact same subjects in every
condition.

24. Type I error is when we reject the null hypothesis when it is true. In other words, we
claim there is a significant effect when there isnt. If we use p<.05 as our alpha level
(i.e., our criterion for determining significance), we are going to make this type of error
5% of the time.
25. Internal validity is the strength of the connection between the iv and the dv. The
more confounds there are in your experiment, the less internal validity you will have. If
you assign participants to groups in a non-random way, you introduce possible confounds
into your experiment, which will threaten the internal validity.
26. Interrater reliability is defined as the correlations between the scores assigned by two
(or more) different raters.
27. In a 2 x 3 design, there are two independent variables, but if you include the
interaction of the two, there are three effects: the two main effect and the interaction. I
gave you points if you chose either two or three. So, most of you got this one right.
28. Another Hawthorne effect question. You just need to know what it is to be able to
answer this one.

29. For this question, you need to know the difference between threats to internal validity
and to ecological validity. If you dont, that may be something you want to review prior
to the final. One threat to internal validity, by the way, is selection-maturation
interaction.
30. Alpha is the probability of making a Type I error and beta is the probability of
making a Type II error.