Chapter 8: Hypothesis Testing Flashcards
hypothesis testing
A statistical method that uses sample data to evaluate a hypothesis about a population.
goal of hypothesis testing
rule out chance (sampling error) as a plausible explanation for the results from a research study.
4 steps of a hypothesis test
- State hypothesis about the population
- Use your hypothesis to predict the characteristics the sample should have.
- Obtain a sample from the population
- Compare data with the hypothesis prediction
if individuals in the treated sample are different than the population there is _____ effect
evidence of an
null hypothesis
states that there is no change in the general population before and after an intervention.
In the context of an experiment, the independent variable had no effect on the dependent variable
alternative hypothesis
states that there is a change in the general population following an intervention.
In the context of an experiment, the independent variable did have an effect on the dependent variable
null distribution
the distribution of the population if there is no effect of treatment
what is the mean of a sample from the null distribution
same as the population mean
if the sample is on the tails of the null distribution, then
there was an effect and we should reject the null hypothesis
H₀
null hypothesis
H1
alternative hypothesis
α level
the criterion used to make a decision about the hypothesis
common alpha levels
α = .01, α = .05 (most used), α = .001
critical region
Outcomes that are very unlikely to occur if the null hypothesis is true.
when do you decide on the alpha level?
before seeing the results
non-directional hypothesis
doesn’t explicitly say if the treatment will increase or decrease the dependent variable
z-values for an alpha level of 0.05
+ 1.96, -1.96
z-values for an alpha level of 0.01
+2.58, -2.58
z-values for an alpha level of 0.001
+3.30, -3.30
rejecting the null hypothesis
treatment had an effect
failing to reject the null hypothesis
treatment did not have an effect
type 1 error
Occurs when the sample data appear to show a treatment effect when there is none. the researcher will reject the null hypothesis and falsely conclude that the treatment had an effect
false positive is another word for a
type 1 error
telling someone who’s not pregnant that they are is a
type 1 error
type 2 error
Occurs when the sample does not appear to have been affected by the treatment when it does. In this case, the researcher will fail to reject the null hypothesis and falsely conclude that the treatment does not have an effect
false negative is another word for a
type 2 error
telling someone who’s pregnant that they are not is a
type 2 error
type 1 errors are represented by
ɑ
type 2 errors are represented by
β
directional tests
include the directional prediction in the statement of the hypotheses and in the location of the critical region
what type of hypothesis test is most common in psych research
nondirectional
Two main differences between directional and non-directional hypothesis tests
- How you write and specify the null and alternative hypothesis
- The critical value associated with your cut-off
5% cut-off with a two-sided test means 2.5 on either side corresponding to a z-score of 1.96 on either side
5% cut-off with a one-sided test means 5.0 on one side, corresponding to a z-score of 1.65
Raw effect size
the effect size in the original unit
cohen’s d formula
d = M-μ/σ
null distribution
distribution if there is no effect
The probability of a type 1 error
Ɑ
probability of a type 2 error
β
power formula
1-β
what is power?
Probability of detecting the effect if it exists
what are typical values for power
0.8 or 0.9
effect size and power
Larger differences between groups or sample and population will have more power
sample size and power
Getting a bigger sample increases power, because it decreases the standard error
a-value and power
Larger alpha values increase power
directional vs. non-directional hypothesis tests and power
One tailed (directional) hypotheses have more power because the sample has to be less extreme for an effect to be found
how can researchers influence power?
- choose a larger a-value
- get a larger sample size
issue with hypothesis tests
they don’t indicate how large the effect is
what cohen’s d value is associated with a small effect
0.2
what cohen’s d value is associated with a medium effect
0.5
what cohen’s d value is associated with a large effect
0.8
